Biopolymers at Interfaces, Second Edition (Surfactant Science)

© 2003 by Marcel Dekker, Inc. Library of Congress Cataloging-in-Publication Data A catalog record for this book is av...

Author: Martin Malmsten

43 downloads 1091 Views 7MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

© 2003 by Marcel Dekker, Inc.

Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress. ISBN: 0-8247-0863-6 This book is printed on acid-free paper. Headquarters Marcel Dekker, Inc. 270 Madison Avenue, New York, NY 10016 tel: 212-696-9000; fax: 212-685-4540 Eastern Hemisphere Distribution Marcel Dekker AG Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland tel: 41-61-260-6300; fax: 41-61-260-6333 World Wide Web http://www.dekker.com The publisher offers discounts on this book when ordered in bulk quantities. For more information, write to Special Sales/Professional Marketing at the headquarters address above. Copyright 䉷 2003 by Marcel Dekker, Inc.

All Rights Reserved.

Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming, and recording, or by any information storage and retrieval system, without permission in writing from the publisher. Current printing (last digit): 10 9 8 7 6 5 4 3

2

1

PRINTED IN THE UNITED STATES OF AMERICA






Preface

It has been a few years since the first edition of Biopolymers at Interfaces was published. Issues relating to the interfacial behavior of biopolymers and notably proteins are as relevant as ever both to understanding biophysical processes and to controlling numerous industrial applications. In recent years, there has been continued progress in our understanding of the basics of biopolymer interfacial behavior, and new experimental methods are being developed and applied for further investigation. Furthermore, the role of biopolymer interfacial behavior in numerous applications is progressively becoming better understood, and at the same time new and exciting application areas are emerging. The second edition of Biopolymers at Interfaces addresses many of these issues. For example, the interesting development of biopolymer-based multilayer structures (with potential applications in, e.g., biosensors) is discussed in the chapters by Schaaf and Voegel (Chapter 13) and Ariga and Lvov (Chapter 14). Also, the progress made in the understanding of protein behavior at the air–water interface, facilitated by developments in experimental techniques, is highlighted and detailed examples are given by Britt et al. (Chapter 16) and Horne and Rodriguez Patino (Chapter 30). Experimental developments are also reviewed by Fromell et al. (calorimetry, Chapter 20), Malmsten (ellipsometry and reflectometry, Chapter 21), Sinner et al. (SPR, Chapter 22), Lu (neutron reflection, Chapter 23), and Griesser et al. (XPS, ToF-SIMS, MALDI-MS, Chapter 24). The section covering different experimental techniques has thus been significantly expanded compared with the first edition. Finally, some fascinating recent developments concerning protein interfacial behavior in microfabricated total analysis systems and microarrays are discussed by De´rand and Malmsten (Chapter 28), and the importance of protein interfacial behavior in the oral cavity (e.g., in relation to biological function) is highlighted by Arnebrant (Chapter 29). Biological function of interfacial protein layers is also the focus of the chapter by Horbett (Chapter 15). I would like to thank all the contributors for their excellent support during the preparation of this edition. Martin Malmsten © 2003 by Marcel Dekker, Inc.

Contents

Preface Contributors Part I.

Concepts

1.

Macromolecular Adsorption: A Brief Introduction Martinus A. Cohen Stuart

2.

Driving Forces for Protein Adsorption at Solid Surfaces Willem Norde

3.

Thermodynamics of Adsorption of Amino Acids, Small Peptides, and Nucleic Acid Components on Silica Adsorbents Vladimir A. Basiuk

4.

Quantitative Modeling of Protein Adsorption Charles M. Roth and Abraham M. Lenhoff

5.

Interfacial Behavior of Protein Mutants and Variants Martin Malmsten, Thomas Arnebrant, and Peter Billsten

6.

Orientation and Activity of Immobilized Antibodies James N. Herron, Hsu-kun Wang, Veˇra Janatova´, Jacob D. Durtschi, Karin Caldwell, Douglas A. Christensen, I-Nan Chang, and Shao-Chie Huang

7.

Interactions Between Surfaces Coated with Carbohydrates, Glycolipids, and Glycoproteins Per M. Claesson

8.

Protein Adsorption Kinetics Jeremy J. Ramsden

9.

Mobility of Biomolecules at Interfaces Robert D. Tilton


vi

Contents

10.

Protein Adsorption in Relation to Solution Association and Aggregation Tommy Nylander

11.

Mechanism of Interfacial Exchange Phenomena for Proteins Adsorbed at Solid–Liquid Interfaces Vincent Ball, Pierre Schaaf, and Jean-Claude Voegel

12.

Interactions Between Proteins and Surfactants at Solid Interfaces Marie Wahlgren, Camilla A.-C. Karlsson, and Stefan Welin-Klintstro¨m

13.

Toward Functionalized Polyelectrolyte Biofilms Pierre Schaaf and Jean-Claude Voegel

14.

Self-Assembly of Functional Protein Multilayers: From Planar Films to Microtemplate Encapsulation Katsuhiko Ariga and Yuri M. Lvov

15.

Biological Activity of Adsorbed Proteins Thomas A. Horbett

16.

Protein Interactions with Monolayers at the Air–Water Interface David W. Britt, G. Jogikalmath, and Vladimir Hlady

Part II.

Methods

17.

Local and Global Optical Spectroscopic Probes of Adsorbed Proteins Vladimir Hlady and Jos Buijs

18.

Proteins on Surfaces: Methodologies for Surface Preparation and Engineering Protein Function Krishnan K. Chittur

19.

Studies on the Conformation of Adsorbed Proteins with the Use of Nanoparticle Technology Peter Billsten, Uno Carlsson, and Hans Elwing

20.

Scanning Calorimetry in Probing the Structural Stability of Proteins at Interfaces Karin Fromell, Shao-Chie Huang, and Karin Caldwell

21.

Ellipsometry and Reflectometry for Studying Protein Adsorption Martin Malmsten

22.

Surface Plasmon Resonance Spectroscopies for Protein Binding Studies at Functionalized Surfaces Eva-Kathrin Sinner, Kazutoshi Kobayashi, Thomas Lehmann, Thomas Neumann, Birgit Prein, Ju¨rgen Ru¨he, Fang Yu, and Wolfgang Knoll

23.

Neutron Reflection Study of Protein Adsorption at the Solid–Solution Interface J. R. Lu


Contents

24.

vii

XPS, ToF-SIMS, and MALDI-MS for Characterizing Adsorbed Protein Films Hans J. Griesser, Sally L. McArthur, Matthew S. Wagner, David G. Castner, Peter Kingshott, and Keith M. McLean

Part III.

Applications

25.

Interaction of Proteins with Polymeric Synthetic Membranes Georges Belfort and Andrew L. Zydney

26.

Protein Adsorption in Intravenous Drug Delivery Martin Malmsten

27.

Control of Protein Adsorption in Solid-Phase Diagnostics and Therapeutics Krister Holmberg and Gerard Quash

28.

Protein Interfacial Behavior in Microfabricated Analysis Systems and Microarrays Helene De´rand and Martin Malmsten

29.

Protein Adsorption in the Oral Environment Thomas Arnebrant

30.

Adsorbed Biopolymers: Behavior in Food Applications David S. Horne and J. M. Rodriguez Patino


Contributors

Katsuhiko Ariga ERATO Nanospace Project, Japan Science and Technology Corporation, Tokyo, Japan Thomas Arnebrant Pharmaceuticals and Food Section, Institute for Surface Chemistry, Stockholm, and Department of Prosthetic Dentistry, Malmo¨ University, Malmo¨, Sweden Vincent Ball* Department of Biophysical Chemistry, Biocenter, University of Basel, Basel, Switzerland Vladimir A. Basiuk Instituto de Ciencias Nucleares, Universidad Nacional Autońoma de Me´xico, Mexico City, Mexico Georges Belfort Howard P. Isermann Chemical Engineering Department, Rensselaer Polytechnic Institute, Troy, New York, U.S.A. Peter Billsten† Department of Applied Physics, IFM, Linko¨ping University, Linko¨ping, Sweden David W. Britt Department of Biological Engineering, Utah State University, Logan, Utah, U.S.A. ˚ ngstrom Laboratory, Division of Ion Physics, Uppsala University, Jos Buijs A Uppsala, Sweden Karin Caldwell Sweden

Center for Surface Biotechnology, Uppsala University, Uppsala,

Current affiliation: * Institute of Chemistry, LIMBO, Strasbourg, France. † Analysis and Formulations, Astra Draco AB, Lund, Sweden.


x

Uno Carlsson Sweden

Contributors

Department of Chemistry, IFM, Linko¨ping University, Linko¨ping,

David G. Castner Departments of Bioengineering and Chemical Engineering, University of Washington, Seattle, Washington, U.S.A. I-Nan Chang Biochemistry Division, Development Center for Biotechnology, Taipei, Taiwan, Republic of China Krishnan K. Chittur Department of Chemical Engineering, University of Alabama in Huntsville, Huntsville, Alabama, U.S.A. Douglas A. Christensen Lake City, Utah, U.S.A.

Department of Bioengineering, University of Utah, Salt

Per M. Claesson Department of Chemistry, Surface Chemistry, Royal Institute of Technology, and Institute for Surface Chemistry, Stockholm, Sweden Martinus A. Cohen Stuart Department of Physical and Colloidal Chemistry, Wageningen Agricultural University, Wageningen, The Netherlands Helene De´rand

Gyros AB, Uppsala, Sweden

Jacob D. Durtschi City, Utah, U.S.A.

Department of Bioengineering, University of Utah, Salt Lake

Hans Elwing Department of General and Marine Microbiology, Go¨teborg University, Go¨teborg, Sweden Karin Fromell Sweden

Center for Surface Biotechnology, Uppsala University, Uppsala,

Hans J. Griesser Ian Wark Research Institute, University of South Australia, Mawson Lakes, Australia James N. Herron Department of Material Science and Engineering, University of Utah, Salt Lake City, Utah, U.S.A. Vladimir Hlady Department of Bioengineering, University of Utah, Salt Lake City, Utah, U.S.A. Krister Holmberg Department of Applied Surface Chemistry, Chalmers University of Technology, Go¨teborg, Sweden Thomas A. Horbett Departments of Bioengineering and Chemical Engineering, University of Washington, Seattle, Washington, U.S.A. David S. Horne

Food Quality Group, Hannah Research Institute, Ayr, Scotland


Contributors

xi

Shao-Chie Huang U.S.A.

Diagnostic Products Corporation, Los Angeles, California,

Veˇra Janatova´* Department of Material Science and Engineering, University of Utah, Salt Lake City, Utah, U.S.A. G. Jogikalmath Department of Material Science and Engineering, University of Utah, Salt Lake City, Utah, U.S.A. Camilla A.-C. Karlsson Lund, Sweden

Department of Food Engineering, University of Lund,

Peter Kingshott kilde, Denmark

The Danish Polymer Center, Risø National Laboratory, Ros-

Wolfgang Knoll Mainz, Germany

Materials Science, Max-Planck Institute for Polymer Research,

Kazutoshi Kobayashi Research and Development Center, Hitachi Chemical Company, Ltd., Ibaraki, Japan Thomas Lehmann New Applications Business Team, Wacker Polymer Systems, Burghausen, Germany Abraham M. Lenhoff Department of Chemical Engineering, University of Delaware, Newark, Delaware, U.S.A. J. R. Lu Department of Physics, University of Manchester Institute of Science and Technology, Manchester, England Yuri M. Lvov Institute for Micromanufacturing, Louisiana Tech University, Ruston, Louisiana, U.S.A. Martin Malmsten Institute for Surface Chemistry, and Department of Chemistry, Surface Chemistry, Royal Institute of Technology, Stockholm, Sweden Sally L. McArthur Department of Bioengineering, University of Washington, Seattle, Washington, U.S.A. Keith M. McLean Department of Molecular Science, Commonwealth Scientific and Industrial Research Organisation, Clayton, Australia Thomas Neumann

Max-Planck Institute for Polymer Research, Mainz, Germany

Willem Norde Department of Physical and Colloid Chemistry, Wageningen Agricultural University, Wageningen, The Netherlands * Present location: Prague, Czech Republic.


xii

Contributors

Tommy Nylander Department of Physical Chemistry 1, Center for Chemistry and Chemical Engineering, University of Lund, Lund, Sweden Birgit Prein Austria

Department of Biochemistry, Graz University of Technology, Graz,

Gerard Quash Oullins, France

Laboratory of Immunochemistry, Faculty of Medicine Lyon-Sud,

Jeremy J. Ramsden Institute of Experimental Medicine, Hungarian Academy of Sciences, Budapest, Hungary J. M. Rodriguez Patino Seville, Seville, Spain

Department of Chemical Engineering, University of

Charles M. Roth Center for Engineering in Medicine, Massachusetts General Hospital, Harvard Medical School, and Shriners Burns Hospital, Boston, Massachusetts, U.S.A. Ju¨rgen Ru¨he Institute for Microsystems Technology, University of Freiburg, Freiburg, Germany Pierre Schaaf

Institut Charles Sadron (CNRS–ULP), Strasbourg, France

Eva-Kathrin Sinner Membrane Biochemistry, Max-Planck Institute for Biochemistry, Martinsried, Germany Robert D. Tilton Colloids, Polymers, and Surfaces Program, Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania, U.S.A. Jean-Claude Voegel Strasbourg, France

Faculte´ de Chirurgie Dentaire, INSERM—Unite´ 424,

Matthew S. Wagner Department of Chemical Engineering, University of Washington, Seattle, Washington, U.S.A. Marie Wahlgren Sweden

Department of Food Technology, University of Lund, Lund,

Stefan Welin-Klintstro¨m Department of Physics and Measurement Technology, Linko¨ping University, Linko¨ping, Sweden Hsu-kun Wang Department of Material Science and Engineering, University of Utah, Salt Lake City, Utah, U.S.A. Fang Yu

Max-Planck Institute for Polymer Research, Mainz, Germany

Andrew L. Zydney Department of Chemical Engineering, University of Delaware, Newark, Delaware, U.S.A. © 2003 by Marcel Dekker, Inc.

1 Macromolecular Adsorption: A Brief Introduction MARTINUS A. COHEN STUART Wageningen Agricultural University, Wageningen, The Netherlands

I.

BIOLOGICAL MACROMOLECULES

The primary structure of many biological macromolecules is simply that of a linear chain of monomers. In this respect, they do not differ from most synthetic polymers. Behind the chemical simplicity, however, rather complex and specific behavior may be hidden. For example, the secondary and tertiary structures of proteins often play an essential role in their specific biological function. For such molecules it may seem impossible to develop a universal description as is now well established for synthetic polymers. However, one may wonder whether this has been attempted. Some of the concepts fruitfully introduced in polymer physics may be quite relevant for understanding the behavior of, say, globular proteins. It is the purpose of this chapter to discuss polymers at interfaces from a physical chemist’s point of view, highlighting processes and central concepts rather than specific systems. Effects of conformational entropy, internal cohesion, electrostatic interactions, and intermolecular interactions on adsorbed layers will be considered, and processes like transport, attachment, unfolding, and trapping will be discussed. As said, most biological polymers are chemically more complex than homopolymers and have a copolymer character that cannot be ignored. However, considering the entire plenitude of phenomena that various classes of copolymers can display would by far exceed the limits of this introductory presentation. We will therefore not explicitly discuss copolymers.

II.

MACROMOLECULES ARE SOFT PARTICLES

Around the bonds that connect the atoms in the main chain of a linear polymer, some rotation is usually possible. This gives the polymer a certain degree of flexibility, which is usually enough to allow the molecule to assume a large number of shapes or conformations. With this large number of conformations, a conformation entropy is associated. In the absence of external forces, the chain will wiggle because of thermal motion and assume an average, approximately spherical coil shape with © 2003 by Marcel Dekker, Inc.

2

Cohen Stuart

a characteristic size R dependent on the number of monomers N per chain (i.e., the length of the chain). This dependence can be generally expressed as a power law: R ⬃ N v. The conformational entropy gives the coil a certain elastic resistance to deformations such as squeezing and stretching. This resilience is rather weak; the molecules can be considered ‘‘soft spheres.’’ The characteristic size R of the coil depends strongly on the solvent [1,2], but it can always be written as a power law of the chain length N: R ⬃ N ␣. Good solvents tend to keep the individual monomers separate. In such a solvent, most shapes permitted by bond rotation are energetically equivalent. If the monomers would not occupy any volume (i.e., the chain is infinitely thin, this is called an ideal chain, analogous to ideal gases), one could describe the chain in terms of a random walk (diffusion-like) process. However, because monomers do have a volume, walks that lead to overlap between monomers must be excluded. The proper description is therefore that of a self-avoiding walk. The characteristic size R of such a structure is known to scale as N 3/5 (␣ = 3/5), and such molecules are denoted as swollen coils. The monomer–monomer repulsion also ensures that the molecules do not cluster but form a homogeneous solution. At higher concentration, however, this repulsion is not enough to keep the coils apart; as soon as they fill the entire solution volume, they begin to interpenetrate, thus forming a transient network. This situation is now commonly referred to as a semidilute solution. In less good solvents, there is, effectively, an attraction between the monomers. As long as this attraction is very weak, the coil may somewhat contract, but it remains swollen. At a certain point, the attraction becomes strong enough to compensate the effect of the excluded volume of the monomers. This is called the ␪ point or ␪ temperature. The behavior of the coil around the ␪ point is equivalent to that of an ideal chain, for which random walk statistics apply and one can prove R ⬃ N 1/2 (␣ = 1/2); this is called a Gaussian coil because the density of monomers inside the coil is a Gaussian function of the distance to the center [2]. Should the solvent become even worse, the entropy can no longer maintain the open, dilute structure of the coil, and the molecule collapses to a compact globule. Inside this globule, the monomer density is essentially constant and therefore the mass must scale as R 3 so that R ⬃ N 1/3 (␷ = 1/3). Of course, monomer–monomer attraction will not only occur between monomers within the same chain, but also between monomers on different chains. Hence, in poor solvents there is a concentration where the molecules accumulate into a dense phase and we have phase separation. As a parameter that describes the strength of the monomer interaction it is customary to use the so-called Flory–Huggins parameter ␹, which is essentially an excess Gibbs energy of monomer–solvent contacts (with respect to monomer– monomer and solvent–solvent contacts) normalized by kT. In good solvents, ␹ is close to zero; at the ␪ point it equals 0.5; and for poor solvents, ␹ > 0.5. In many texts, the parameter ␷ = 1 ⫺ 2␹ is used. It measures the strength of the monomer– monomer interaction relative to that at the ␪ point and is called the excluded volume parameter. Its value is zero at the ␪ point, positive in better solvents, and negative in worse solvents (Fig. 1). In poor solvents, where the molecules collapse into globules, deformation is not only counteracted by the conformational entropy, but also by the fact that the number of (unfavorable) monomer–solvent contacts increases at the expense of (fa© 2003 by Marcel Dekker, Inc.

Macromolecular Adsorption

3

FIG. 1 Polymer molecules in solution under (a) good, (b) ␪, and (c) poor solvent conditions.

vorable) monomer–monomer contacts, just as for a droplet of an inmiscible liquid in some liquid medium. One may therefore assign a certain interfacial tension ␥ to the molecules of order kT␷ 2. This means that collapsed macromolecules are also deformable, but to an extent dictated by the surface tension (␷ 2) rather than the conformational entropy. in other words, they become ‘‘harder’’ as 兩␷ 兩 increases [3]. As is well known, protein molecules are rather compact. This compact structure is stabilized by hydrophobic attractions between lipophilic amino acids which tend to accumulate inside the globular structure of the molecule because water is a poor solvent for them. In this sense, there is a clear parallel with simple homopolymers in a poor solvent. As one approaches the temperature where the hydrophobic interactions become too weak to keep the globule together, the particles become ‘‘softer’’ (get lower surface tension), and unfolding becomes increasingly easier. Therefore one can consider the denaturation temperature of the protein as a kind of ␪ temperature. Of course, the important difference between proteins and simple polymers in a poor solvent is that the former are copolymers; they have sufficient hydrophilic monomers on their external surface to suppress the accumulation of molecules into a macroscopic phase. A detailed discussion of protein structure can be found in Chapter 2.

III.

ADSORPTION AND PARTICLE DEFORMATION

Adsorption of a polymer occurs whenever its monomers are attracted sufficiently strongly by a surface. Therefore, on the one hand, the molecules attempt to maximize their favorable monomer–surface interactions by spreading out to a large extent. For each polymer–surface contact established, a solvent molecule is detached and the net Gibbs energy of contact formation is therefore up ⫺ us. For convenience, the parameter ␹s = (up ⫺ us)/kT has been introduced as a measure of adsorption affinity. On the other hand, there is a penalty in the form of the resistance to deformation. For the Gaussian or swollen coil, this resistance is weak. In the case of weak attraction, the polymer is deformed into a ‘‘pancake’’ of thickness D, where D scales as (␧/kT)⫺␣/(1⫺␣), where ␧ is the adsorption energy per monomer [1]. For the Gaussian coil, one may argue that the Gibbs energy increase due to entropy loss (⫺TS ) upon adsorption is, per monomer unit, of order ln(␻b/␻s), where ␻b and ␻s are the orientational degrees of freedom per monomer in the bulk (b) and in the surface (s), respectively. This entropic contribution is small, typically around 0.5 kT, and must be balanced by an equal adsorption energy ␧ for appreciable adsorption to occur. Values of 1 kT for ␧ are quite common; this is enough to completely squeeze the chains into an essentially flat structure (Fig. 2a) [4]. © 2003 by Marcel Dekker, Inc.

4

Cohen Stuart

FIG. 2 Polymers on a surface: (a,b) good solvent conditions; (c) poor solvent conditions.

Hence, already at moderate adsorption strength per monomer, large polymers can make an appreciable number of contacts with the surface and will, therefore, adhere tenaciously. As a consequence, the surface will be fully covered before any free polymer in the equilibrium solution is detected. In other words, the adsorption has a strongly high-affinity character. When the molecules start to crowd on the surface, they begin to interpenetrate, thus forming a more or less homogeneous layer on the surface with a high monomer concentration (Fig. 2b). Since in good solvents, the monomers repel each other, the Gibbs energy of adsorption increases in this stage, eventually bringing the system to equilibrium. The homogeneity of the adsorbed layer allows us to consider it as a polymer solution with a concentration which only varies normal to the surface, the so-called density profile [4]. In Section V I explain how one can calculate this profile theoretically from known solution properties of the polymer and from the adsorption energy. For the globular polymer molecule (Fig. 2c), deformation leads to exposure of more monomers to the solvent. In other words, work must be done against the interfacial tension. Therefore, globular polymers adsorb in a way analogous to the spreading of liquid droplets [5]. The driving force is, again, given by the low polymer–substrate interfacial tension (proportional to ␹s), but the resistance is a measure of the monomer–monomer attraction as given by ␹. Large deformation will occur as soon as ␹x = ␹ 2. For a strongly deformed polymer, the monomer–monomer contacts are so few that one can no longer speak of a surface tension; this occurs when the thickness becomes of order ␰t, the thermal blob size given by ⫺1/v [3,5].

IV.

FORMATION OF AN ADSORBED LAYER

When a bare surface is exposed to a solution of adsorbing macromolecules, an adsorbed layer will form. Each adsorbing molecule must pass through the following steps: 1. 2. 3.

Transport toward the surface of convection and diffusion Attachment Spreading

In the following, we shall assume that the formation of the polymer layer takes place under well-defined hydrodynamic conditions, so that the first step (transport) can be treated simply by solving the convective diffusion equation. Analysis shows that most of the process takes place under steady-state conditions where a fixed concentration gradient drives the process. This leads to an adsorption rate [6] © 2003 by Marcel Dekker, Inc.


d⌫ = k(cb ⫺ cs) dt

5

(1)

where cs is the concentration of polymer in the immediate neighborhood of the surface, often referred to as the subsurface concentration, cb is the bulk concentration, and k is a rate coefficient depending on diffusion coefficient and hydrodynamic conditions. Expressions for k in various flow geometries have been given by Adamczyk et al. [7]. The value of the subsurface concentration cs depends on what happens near the surface, i.e., on the attachment process. In general, one can treat this as a first-order reaction. For the forward rate,

冏

d⌫ dt

= Kcs

(2)

⫹

Of course, the rate constant K will increase with increasing coverage ⌫, because fewer and fewer surface sites are available for the polymer to attach to. However, even on a bare surface, it may be lowered by some repulsive barrier (e.g., electrostatic repulsion). As the coverage increases, so does the rate of desorption. Eventually, the system should come to its equilibrium condition, and the rates of adsorption and desorption should become equal. This implies for the net attachment rate, i.e., the sum of forward and backward rates: d⌫ = K(cs ⫺ ceq) dt

(3)

where ceq is the equilibrium concentration corresponding to the value of ⌫ at some moment during adsorption. The function ceq(⌫) is the inverse adsorption isotherm. Combining Eqs. (1), (2), and (3), one obtains d⌫ cb ⫺ ceq = dt 1/k ⫹ 1/K

(4)

Provided the attachment barrier is negligible (K ⫺1 = 0) and the adsorbed layer entirely relaxed, one can use this equation to deduce the adsorption rate from the equilibrium isotherm and vice versa. For example, the high-affinity adsorption isotherm so characteristic for most polymers implies that ceq ⯝ 0 up to almost saturation, so that d⌫/dt = kcb, and ⌫ must increase linearly in time with a slope k up to almost saturation; this is easily verified in a well-controlled experiment [8]. Since the rate of adsorption cannot increase beyond kcb, this is called the limiting flux J0. Lower rates are indicative for adsorption barriers. One can also use the equation to describe desorption into pure solvent by putting cb = 0 so that d⌫/dt = ⫺kceq(⌫) [9]. In order to develop this further and predict (starting from the same assumptions) adsorption and desorption rates, the function ceq(⌫) and, hence, an equilibrium adsorption theory is required. This will be considered in the following section. V.

THE HOMOGENEOUS LAYER IN EQUILIBRIUM

Developing a general theory for an arbitrary kind of polymer on an arbitrary surface is of course a hopeless task. We therefore shall have to limit ourselves to a tractable © 2003 by Marcel Dekker, Inc.

6

Cohen Stuart

case, which nevertheless brings out essential features. We shall consider an ideally planar and homogeneous surface in contact with a solution of a simple, uncharged homopolymer. The polymer is homodisperse: all the polymer molecules have the same molar mass. (This limitation can be overcome, however [10].) The monomer units interact with sites on the surface with a net energy ⫺kT␹s which may be negative (attraction) or positive (repulsion); see Section III. As pointed out, polymers in ␪ or good solvents form a layer of interpenetrated molecules, the density ␾ of which varies only normal to the surface. We now shall try to calculate this density profile ␾(z). One particularly transparent way to arrive at a theoretical description is the meanfield approach. On a crowded surface with many polymers, there are many interactions between the chains. In any statistical theory of adsorption, all these interactions have to be considered. However, this is obviously very complicated. In a mean-field theory, these interactions are replaced by interactions between a single chain and an average environment, called the field, that originates from the presence of all the other chains. Hence, each monomer at a distance z from the surface just ‘‘feels’’ a potential (Gibbs energy) given by the average density at and around z. This reduces the problem to one of a chain in an external, z-dependent field. Of course, the strength of the field at z is a function of the monomer density at z. To a reasonable approximation, one can ignore the interactions of the chain with itself. We then have the case of a Gaussian chain in an external field, which can be treated with random walk statistics and leads to an expression for the density profile and, hence, to a field. Of course, this field must be the same as that which was initially imposed; it must be self-consistent, hence the term self-consistent field theory [4,11,12]. The first step in the calculation of ␾ at position z is to realize that a monomer finding itself at z [so that it contributes to ␾(z)] must be somewhere in a polymer chain. It therefore has to be the end point of two random walks (in the field), starting from either end of the chain, and ending exactly at z. Hence, we can write ␾(z) in terms of the product of (1) the end point probabilities g(z, s) and g(z, N ⫺ s) (with respect to the same probabilities in the solution far from the wall) for the two walks that make up a chain of length N [namely, s steps from one end and (N ⫺ s) steps from the other] and (2) the density of s-monomers in the bulk solution ␾b/N [4]:

␾(z, s) =

␾b g(z, s)g(z, N ⫺ s) N

(5)

We now have to find the function g(z, s) by considering the random walk in a field problem. This is equivalent to solving the diffusion equation

␭

d 2g dg = ug ⫹ dz2 ds

(6)

where ␭ is the fraction of bonds normal to the interface and ug represents the effect of the external field u on g. Equations of this type occur frequently. One example is the motion of elementary particles as described by Schro¨dinger’s equation. The general solution is written as g(z, s) =

冘

␧ si

gi(z)

(7)

i

where ␧i is the eigenvalue of the ith term. Note that the solution is factorized into © 2003 by Marcel Dekker, Inc.


7

z- and s-dependent parts. Short random walks will be rather different from long ones, and therefore g(z, s) and g(z, N ⫺ s) are also different for small values of s or N ⫺ s, i.e., near the end points. However, as the walks become longer, they will visit the same z again and again, and eventually the z-dependence of g becomes more or less independent of s. In terms of the Schro¨dinger equation, we expect stationary solutions for long times and a spectrum of well-separated stationary eigenfunctions. The assumption that is therefore usually made at this point is that only the first term in the series of Eq. (7) is important; this is equivalent to supposing that the system is in its ‘‘ground state’’ (this is called the ground state approximation, or GSA) [4]. Hence, ␧s

g(z, s) = g(z)e

(8)

Here, ␧ can be seen as the free energy per monomer in the ground state. Using this we can rewrite Eq. (5) as

␾(z, s) =

␾b 2 ␧N g e N

(9)

Summing this over s gives

␾(z) = N␾(z, s) = ␾bg 2e␧N

(10)

We note that in the solution far from the wall we must have g(z) = 1, giving

␧=⫺

ln(␾b) N

(11)

The ‘‘potential’’ u can be considered as the extra reversible work needed to bring one monomer from the solution to the adsorbed layer, which in good solvents is well approximated by ␾(z). Inserting this into Eq. (6) one obtains, for the case of an attracting wall [and using the boundary condition that ␾(z) has a negative slope at the wall],

␾(z) =

␧ sinh2(z ⫹ D/d)

(12)

where D is a constant related to the adsorption energy per monomer, and d equals

兹␭/␧ [13]. For the adsorption isotherm in the region of the semiplateau, one deduces by integration: ⌫=

2␭ d

再冉冊冎 coth

D d⫺1

⬇ 2␭

冉冊 1 1 ⫺ D d

(13)

Note that the adsorbed amount is linear in ␧, hence to [ln(␾b)/N]1/2; it is not really constant but increases very slowly with the polymer concentration. This modest increase in ⌫ comes from the growth of the dilute periphery of the layer with concentration and with molecular weight. Of course, the approximations introduced can be avoided if one solves the basic equations numerically. Then end effects and short chains are no longer neglected. This has been done in great detail by Scheutjens and Fleer [11]. Further analysis has shown that good agreement between numerical results and analytical theory can be obtained when the expansion Eq. (7) is not truncated after the first term but after the second term; there is a second function g2(z) which has © 2003 by Marcel Dekker, Inc.

8

Cohen Stuart

almost the same eigenvalue as g1 but describes walks that do not touch the wall. In quantum mechanical terms, the spectrum is degenerate. Just as g21 corresponds to chains which have at least two attached monomers (thereby forming loops), g22 relates to conformations that do not touch the wall (free chains), and the cross product g1 g2 to chains that start on the surface and end in the solution (tails). Taking these three contributions into account, excellent agreement between numerical and analytical solutions can be established [4,13–15].

VI.

ADSORPTION AND DESORPTION RATES

The function ⌫(ceq) being available, we can invert it to ceq(⌫) and insert it into Eq. (4). One sees immediately that, provided k is a constant and other resistances do not occur (K ⫺1 = 0), we obtain a straightforward differential equation for ⌫(t). As pointed out, ⌫ is predicted to increase proportionally to t, the slope being given by k. As soon as the semiplateau is reached, ceq increases very rapidly (roughly as e N(⌫ ⫺ ⌫ov), where ⌫ov is the coverage where molecules start to overlap), cb ⫺ ceq drops rapidly, and d⌫/dt becomes very small. Hence ⌫(t) has a second part where it rises very slowly (Fig. 3). In the same way, we can analyze the case of desorption. Now, cb = 0 is imposed. Initially ⌫ decreases, as ceq has the finite value c0, but as ⌫ decreases, ceq must become very small because of the high-affinity character of ⌫(c) (Fig. 3). Hence, d⌫/dt also becomes very small. We can now use Eq. (12); to a very good approximation, ⌫ ⬃ log c. If this is the case, one can show that ⌫ must decay logarithmically [9]: ⌫ = ⌫0

冋

冉冊册

1 ⫺ p ln 1 ⫹

t ␶

(14)

where p is a small coefficient of order 1/N and ␶ equals p/kc0 (Fig. 4). This result teaches us that per decade in time a constant (small) amount p⌫0 is desorbed, say 1% of the initial value. Hence, the time needed for desorption increases exponentially with the amount to be desorbed. It follows that, in a practical sense, long chains desorb too slowly for any rate to be measured, which is in agreement with the experimental observation that adsorbed polymer layers cannot simply be removed

FIG. 3 (a) Adsorption isotherm ⌫(c); (b) corresponding kinetic curve ⌫(t).



9

FIG. 4 (a) Adsorption isotherm ⌫(log c); (b) corresponding desorption curve. Note that the slope of ⌫(log c) is the same as the one for ⌫(log t) but with a minus sign.

by rinsing with the solvent they were prepared from [9]. Yet one should realize that this does not imply that densities and conformations at the surface cannot relax!

VII.

STRUCTURE OF ADSORBED LAYERS

As pointed out, isolated random coil polymers tend to become strongly compressed normal to the surface. However, as soon as crowding occurs, the driving force for the spreading decreases because entirely free surface is no longer available. More polymer can adsorb, but the molecules must protrude into the solution and form loops and loose ends (‘‘tails’’). Hence, the dilute periphery of the adsorbed layer grows in the normal direction. The protrusions, in particular the tails, have a pronounced effect on solvent flow along the surface. Despite their very low density they are able to suppress flow velocities near the surface very strongly, just as a few trees on an open field break the wind [4,16]. As a consequence, the surface apparently bears a considerable stagnant layer of solvent, the thickness of which is called the hydrodynamic layer thickness ␦h. Measurements and calculations show that ␦h is very small at low coverage (chains spread out) but increases enormously as soon as saturation is approached, and that the maximum layer thickness observed becomes strongly dependent on molecular weight. This is clearly shown by the (theoretical) curves given in Fig. 5 [16]. In some systems, particularly for good solvents, a dependency of ⌫ on molecular weight could not be detected, yet ␦ increased strongly with increasing M; the few dominating tails hardly contribute to the adsorbed mass [17]. This corresponds quite well with what we see in Fig. 5 for good solvents: ␦ may increase very strongly with r while ␪ is virtually constant. Turning this around, one concludes that very minor desorption of molecules from a saturated layer must show up in a strong decrease of ␦h; this idea has been exploited to verify the desorption rate equation [9]. Data are shown in Fig. 6 for static thickness measurements (at finite cb) and for the dynamic case (cb = 0). © 2003 by Marcel Dekker, Inc.

10

Cohen Stuart

FIG. 5 Hydrodynamic layer thickness ␦h as a function of coverage ␪ for various molar masses (here indicated by the number of chain segments r). The curves on the left are for ␹ = 0, those on the right for ␹ = 0.5, a ␪ solvent. These curves are obtained by a theoretical based on flow perturbation in a (theoretically calculated) density profile [4].

VIII.

POLYMER MIXTURES: PREFERENCE AND EXCHANGE

It is often stated that polymers adsorb irreversibly, implying that a molecule, once adsorbed, cannot leave the surface anymore. This is a very misleading idea. It has been observed for many systems that an adsorbed polymer molecule can readily desorb if another molecule (even when identical to the leaving one) takes its place

FIG. 6 (a) Decrease of layer thickness as a function of log t during a kinetic desorption experiment (the layer is continuously flushed with pure solvent); (b) increase of layer thickness as a function of log c in a static thickness measurement (fixed c for each point). Note the similarity in slope and shape between ␦h(log t) and ␦h(log c), just as between ⌫(log t) and ⌫(log c) in Fig. 4.



11

so that ⌫ does not need to decrease. In other words, exchange with molecules in solution occurs frequently. This may have various consequences. When the solution contains more than one solute component, exchange may lead to a change in the composition of the adsorbed layer. Most commercial polymers are far from pure components, as they contain a large variation in chain lengths so that one may anticipate shifts in the average chain length on the surface. Such polydispersity effects have indeed been observed in many systems [4,18,19]. In dilute solutions, there is a clear driving force for preferential adsorption of long chains. Each long chain (of length N) that adsorbs at the expense of an equal mass of short chains (of length n) liberates on average N/n chains, which leads to a small net gain in translational entropy of order kT(N/n ⫺ 1). Since exchange can even be observed in the absence of a net driving force, we should not be surprised to see exchange occur between chemically identical short and long chains. For an entire distribution of chain lengths (as is the usual case) one can show that the surface chain length shifts gradually to higher values as the polymer concentration in solution is increased; the amount of available surface area as compared to the solution volume is then a crucial parameter, determining the final distribution of each fraction in the polymer sample over the solution and the surface [10,20]. How this exchange process proceeds and what its rate is are discussed in the following section. Chemically different polymer molecules often have an even more pronounced tendency for exchange. In this case the driving force is a lowering of the Gibbs energy of the interfacial contact. Since the adsorbed mass in equilibrium depends only logarithmically on the bulk concentration of a component, whereas the polymersurface interaction is linear in the surface composition, a large concentration difference is needed to offset a difference in surface energy [4,21]. Therefore, simple experiments of sequential adsorption may quickly tell which of a given monomer adsorbs more strongly. The limiting case of exchange between chemically different species is displacement of a polymer by a small molecule. This may even occur when the small displacer molecule adsorbs just a bit stronger than the monomer unit but is present in large amounts [22–24]. A simple case is the adsorption of a polymer from its own ‘‘monomer.’’ Theory predicts that this will not occur as there is no change in surface contact energy, hence no driving force for spreading and adsorption. Diluting the displacer will eventually restore the situation of adsorption; this happens at a sharply defined critical displacer concentration. Thermodynamic analysis allows us to determine the adsorption energy per monomer unit from this critical displacer concentration [22,23]. Occurrence of exchange is also well documented for proteins. When a solution of a protein is exposed to a surface, adsorbed molecules may again spontaneously exchange with molecules from the solution. It has been found for some cases that as a result of adsorption followed by desorption one finds molecules in the solution which have undergone irreversible changes in conformation. Exchange between different proteins occurs frequently in mixtures and is known as the Vroman effect [25]. IX.

SPREADING: EXPERIMENTAL EVIDENCE

In Section III we argued that macromolecular adsorption is always accompanied by deformation or spreading of the molecules. In our analysis of the rate of the ad© 2003 by Marcel Dekker, Inc.

12

Cohen Stuart

sorption process we totally ignored this spreading process. In other words, we considered adsorption processes where spreading occurs very fast as compared to the experimental time scale. As a result, no information on the rate of the spreading process could be obtained. What if this is not so? Qualitatively, one consequence is immediately seen. If the molecules need a certain time ␶s to spread, we can compare that to the time ␶d required to deposit molecules to the surface. If the deposition is slow (␶s < ␶d), each molecule can unfold and spread before it is surrounded by other molecules. The adsorbed amount will correspond to a thin layer and therefore be low. In the reverse case—fast supply (␶s > ␶d)—the molecules are enclosed by neighbors before they had time to unfold. The layer will therefore be thicker and ⌫ higher. The two cases are sketched in Fig. 7. In order to develop this more quantitatively, we need to specify the occupied area as a function of time. Let us suppose that one single time scale ␶s dominates the spreading process, i.e., a = a0 ⫹ ␣(1 ⫺ e⫺t/␶s)

(15)

where a0 is the size of the molecule upon first attachment and ␣ denotes the extra surface occupied at full spreading. In addition, we need to specify the relation between the rate of attachment and the coverage. We simply suppose that this rate is proportional to the unoccupied area ␤. We now have two parallel processes, supply and spreading, which are coupled by the instantaneous value of ␤. This leads to a differential equation for the adsorbed mass versus time, which now becomes a function of the limiting flux J0 [26,27]. If no desorption occurs (so that spreading stops as soon as ␤ = 0), one gets a family of ⌫(t) curves with increasing saturation plateaus as J0 increases. This allows us to determine the spreading rate constant. An experimental example is given in Fig. 8 for the adsorption of the protein savinase [27]. Desorption may also occur, in which case some molecules may increase their occupied area at the expense of others; this could be called ‘‘competitive spreading.’’ If this occurs, one expects that at high supply rates many molecules are packed on the surface. As they begin to spread, some molecules are ‘‘kicked off’’ the surface and the coverage decreases spontaneously; the ⌫(t) curve has an overshoot. At lower supply rates, the spreading process can keep up with the supply, and the overshoot disappears. Such behavior is seen, e.g., for the protein savinase on silica (Fig. 9) [28]. Effects as discussed here occur frequently for proteins but not for most of the swollen polymers, although one case for a polyelectrolyte has been reported. This seems to indicate that the spreading rate is strongly decreased by the high density in the protein globules, probably as a result of strong friction within the molecule

FIG. 7 Adsorbed layers obtained at (a) fast and (b) slow spreading.



13

FIG. 8 Adsorption kinetics of savinase at various concentrations, hence various supply rates (limiting flux) J0 = kcb. The time axis has been renormalized by J0 to make the initial slopes all equal to unity [27].

as it is being deformed. Swollen polymers do not have much internal friction. However, their spreading rate may be slowed down because monomers must detach and reattach to the surface, and this requires an activation energy which is equivalent to friction with the substrate. Alternatively, competition with surrounding molecules may play a role. That surrounding molecules do so is deduced from exchange experiments between two polymeric species where the detected rate of desorption of one species reveals the rate of spreading of the other (Fig. 10) [29,30]. Most observations have shown that this process has two stages: (1) the invading chain attaches rapidly to the surface and (2) the displaced polymer molecules leave the surface more slowly. In fact, in many cases one can stop the supply of displacing polymer without inter-

FIG. 9 Kinetic curve for the adsorption of savinase, showing an overshoot [28].


14

Cohen Stuart

FIG. 10 Desorption of molecules induced by a spreading process; the detected rate of desorption reveals the rate of spreading.

rupting the desorption process; the displacer is already ‘‘stored’’ on the surface. In our view, desorption then follows as a result of competitive spreading, where the invading chain increases its occupied area at the expense of the leaving species. Detailed studies on several systems have shown that the rate of competitive spreading depends on (1) stiffness of the invading chain and (2) its net adsorption energy. This seems plausible, as both increase the friction between the polymer and the substrate.

X.

TRAPPING

Granick [31] has emphasized the importance of topological effects in dense layers of flexible polymers. Just as in concentrated solutions, long chains tend to become entangled, and this hampers translational diffusion very strongly. Hence, instead of displacement by spreading (as described above) an incoming polymer can also trap chains that are already on the surface. The trapped chain must escape from ‘‘under’’ the invading chain which may involve a kind of slithering motion (reptation) that could be very slow indeed. This trapping phenomenon has been investigated in more detail with a system consisting of two rather similar polysaccharides. These were first sequentially supplied to the surface in order to allow displacement and/or trapping to occur. Then the solvent was changed such that the first polymer was encouraged to leave the substrate. Desorption was allowed to take place. Finally, the composition of the adsorbed layer was analyzed. The experiment was repeated for several lengths of the displacer polymer. Surprisingly, both displacement and trapping were observed. When the second polymer was shorter than the first one, displacement did not occur, but substantial trapping was seen. Increasing the displacer’s chain length led to substantial displacement (up to about 80%) and much less trapping (Fig. 11) [32].

XI.

POLYELECTROLYTES AND POLYAMPHOLYTES

So far we have not said anything about electrostatic effects. Yet many water-soluble macromolecules, particularly those of biological origin, carry charges. A simple case is that of a water-soluble homopolymer with a fixed fraction of equally charged monomer. If these changes arise from strong dissociating groups, they will remain © 2003 by Marcel Dekker, Inc.


15

FIG. 11 Fraction of initially adsorbed polymer that is displaced, desorbed, and trapped as a function of displacer/trapper chain length. System: carboxypullulan adsorbed on polystyrene from water, trapped by uncharged pullulan [32].

charged over the entire pH range; this is called a quenched polyelectrolyte. The opposite case, where the charge arises from reversible proton transfer and is thus pH dependent, is the annealed case [33]. Another case of importance, certainly in the context of biological matter, is that of polyampholytes, which carry both positive and negative groups. Polyelectrolytes tend to swell due to essentially two electrostatic effects: the chains become locally more rigid and the effective excluded volume increases. Very long polymers will still have a swollen coil shape, but short ones may eventually become rodlike. Polymers in poor solvents, which would be collapsed in their neutral state, may swell dramatically as soon as they acquire enough charges; there are welldocumented cases for annealed polyelectrolytes which behave in this way (e.g., polymethacrylic acid) [34]. Polyampholytes behave differently. If their net charge is around zero, they tend to contract, because the formation of ion pairs between positive and negative groups leads again to an increase of entropy which drives the association. Synthetic polyampholytes therefore tend to be poorly soluble. Adding salt may improve the solubility and narrow the pH range in which there is phase separation (‘‘salting in’’) because the entropy gain of the ions becomes less. Hence, it may be misleading to consider only the net charge on a molecule. At larger distances (of order ␬⫺1, the Debye length) the local variations in charge become ‘‘invisible’’ and only the net charge is important. Surfaces may be classified in roughly the same way. Many inorganic surfaces consist of metal (hydr)oxides and can be considered as annealed surfaces that get their charge (density ␴s) from reversible proton transfer reactions. Various insoluble salts behave likewise, but they get their charge from other adsorbing ions. Most of these annealed surfaces are amphoteric and have a point where their net charge is © 2003 by Marcel Dekker, Inc.

16

Cohen Stuart

zero; this is the point of zero charge (pzc) [35]. Quenched surfaces are obtained if strongly dissociating groups are chemically attached to a surface; a typical example is the surface of polymer latex particles with sulfate groups. The behavior of polyelectrolytes and polyampholytes is complicated because both the electrostatic and the nonelectrostatic (short range) interactions may contribute, and all sorts of patterns may arise. In order to appreciate the role of electrostatics, we first discuss pure electrosorption of a simple polyelectrolyte on a surface with opposite, quenched charge, from a solution of very low ionic strength [4]. In its neutral state, the polymer experiences no attraction from this surface, and adsorption does not take place. A small amount of charge suffices to induce substantial adsorption up to a point where the charge on the surface and that on the polymer are just equal. Beyond that charge compensation point, only little more charged polymer can adsorb (for entropic reasons the net charge of surface plus polymer need not be zero) up to a certain extent of overcompensation [36,37]. Beyond this value of ␴p the adsorption process can be considered as a kind of ion exchange process, where the macroion takes the place of many small counterions at the surface [38]. Clearly, the driving force for this process is the large entropy gain (at low ionic strength) due to the liberation of the small ions. We should therefore expect that the adsorbed amount comes close to charge stoichiometry. Hence, if the charge density ␴p on the polymer increases further, less and less polymer is needed to reach the same overcompensation; the adsorption decreases with increasing ␴p. A curve of ⌫ versus ␴p is obtained with a pronounced maximum located at the compensation point (Fig. 12) [38]. This pattern is rather general and occurs for many polyelectrolytes. If there are additional, nonelectrostatic interactions contributing to the adsorption, the adsorbed amounts at low ␴p will be increased, particularly those around ␴p = 0. An experimental case of pure electrosorption (for an annealed polyelectrolyte) has been reported (Fig. 13) [39]. In this case, the amphoteric nature of the annealed surface played a crucial role. On the basis of net charge, electrosorption would not be expected to occur beyond the pzc of the surface, where polymer and surface carry the same charge, because there would be no net attraction between the polymer and

FIG. 12 Adsorption ␪ ex of a polyelectrolyte on a surface with fixed charge as a function of polymer charge density ␴p [38].



17

FIG. 13 Adsorption versus pH of PDMAEMA on TiO2 at two different ionic strengths [39].

the surface. Yet, adsorption occurs well into that range. This can only mean that the polymer is able to find and attach to sites of opposite charge; it sees the heterogeneity of the surface. That nonelectrostatic forces might be responsible for this effect could be excluded because the polymer was unable to adsorb in its uncharged form. When we have electrosorption of a quenched polyelectrolyte on an annealed surface, the polymer structure remains unaffected by pH. The system then behaves as a capacitor with a fixed capacitance so that charge and potential are linearly related as are (by Nernst’s law) potential and pH. As a result, the adsorption is simply linear in pH [39]. In many cases, both polymer and surface can adjust their charge. Provided the pzc of the surface is well separated from the pK of the polyelectrolyte, we can observe a combined pattern [40]. At this point, we should consider the question to what extent polyelectrolyte adsorption is reversible. Adsorption of polyelectrolytes is a self-killing process; the accumulation of charged polymer leads to a high electrostatic potential which will repel new incoming molecules. Even if these molecules could anchor under strong short-range interactions (complexation, hydrophobic interaction), they may not be able to reach the surface. If this occurs, the adsorbed amount is kinetically limited, rather than the outcome of a free energy balance. Lowering ␴p (pH shift) or adding salt may then lower the kinetic barrier and promote adsorption, but subsequently restoring the original conditions does not restore the corresponding adsorption. In other words, adsorption hysteresis occurs under cycling the pH or the ionic strength. This is analogous to the aggregation of charged colloidal particles, which can be very effectively enhanced by adding salt or bringing the system into the pzc but often cannot be undone by restoring the conditions under which the dispersion was stable. Kinetic barriers due to electrostatic repulsion for the flexible polyelectrolytes were recently considered theoretically [41]; examples of experimental systems are numerous in the literature. We give one example in Fig. 14. Polyampholytes are attracted to almost any charged surface. As with the amphoteric surface, they can find (short-range) electrostatic attraction even when the net surface charge is repulsive. This has been shown theoretically and experimentally for synthetic polyampholytes [42,43]. Should nonelectrostatic interactions also be present, then adsorption is even more likely to occur, unless a very strong barrier © 2003 by Marcel Dekker, Inc.

18

Cohen Stuart

FIG. 14 Polyelectrolyte adsorption hysteresis upon cycling pH: adsorption of carboxy methyl cellulose on titanium dioxide (TiO2, rutile) particles from aqueous electrolyte solution. Closed circles: adsorption was measured at fixed pH; open circles: adsorption took place at low pH (2.8 and 3.9, respectively) but was determined after increasing the pH [40].

prohibits the attachment of the polymer. Hence, proteins in their pzc are likely to adsorb to almost any substrate and certainly to hydrophobic substrates. Adding electrolyte will reduce all effects due to charges and therefore we should expect to eventually obtain a case of pseudoneutrality [4]. How adsorption depends on ionic strength therefore depends in the first place on the level of adsorption in the absence of any charge. If this is high, one may anticipate salt to enhance adsorption; if this is low or zero, then one must expect lower adsorption [4,37]. In some subtle cases there is a weak maximum because screening of lateral interactions occurs before extensive ion competition for surface sites comes into play [39,44].

XII.

CONCLUDING REMARKS

The adsorption process of a macromolecule is a complex process in which transport in solution, attachment barriers, and shape relaxations together determine the final outcome. In particular, shape relaxations are difficult to study directly, but recent research has shed much light on the factors that control the rate of these surface processes. With the help of these insights, it may soon become feasible to construct a theory for the adsorption of biopolymers.

REFERENCES 1. 2. 3.

P. G. de Gennes, Scaling Concepts in Polymer Physics. Cornell University Press, Ithaca, NY, 1979. H. Yamakawa, Modern Theory of Polymer Solutions. Harper & Row, 1971. I. M. Lifschitz, A. Yu. Grosberg, and A. R. Khoklov, Structure of a polymer globule formed by saturating bonds. Zh. Exp. Teor. Fyz. 71:1634–1643 (1976).


Macromolecular Adsorption 4. 5. 6. 7. 8. 9. 10.

11.

12.

13. 14. 15. 16.

17.

18.

19.

20. 21. 22.

23.

24.

19

G. J. Fleer, M. A. Cohen Stuart, J. M. H. M. Scheutjens, T. Cosgrove, and B. Vincent, Polymers at Interfaces. Chapman & Hall, London, 1993. A. Johner and J. F. Joanny, Polymer adsorption in a poor solvent. J. Phys. II 1:181–194 (1991). M. A. Cohen Stuart and G. J. Fleer, Adsorbed polymers in nonequilibrium situations. Ann. Rev. Mat. Sci. 26:463–500 (1996). Z. Adamczyk, T. Dabros, J. Czarnecki, and T. G. M. van de Ven, Particle transfer to solid surfaces. Adv. Colloid Interface Sci. 19:183–252 (1983). J. C. Dijt, M. A. Cohen Stuart, J. E. Hofman, and G. J. Fleer, Kinetics of polymer adsorption in stagnation point flow. Colloids Surfaces 51:141–158 (1990). J. C. Dijt, M. A. Cohen Stuart, and G. J. Fleer, Kinetics of polymer adsorption and desorption in capillary flow. Macromolecules 25:5416–5423 (1992). G. J. Fleer, Multicomponent polymer adsorption. A useful approximation based upon ground-state solutions. Colloids Surf. A: Physicochem. Eng. Aspects 104:271–284 (1995). J. M. H. M. Scheutjens and G. J. Fleer, Statistical theory of the adsorption of chain molecules. I: Partition function, segment density distribution and adsorption isotherms. J. Phys. Chem. 83:1619–1635 (1979); II: Train, loop, tail size distribution 84:178–190 (1980). P. G. de Gennes, Polymer solutions near an interface. I: Adsorption and depletion layers. Macromolecules 14:1637–1644 (1981); II: Interaction between two plates carrying adsorbed polymer layers 15:492–500 (1982). G. J. Fleer, Ground state description of the adsorption of homodisperse and polydisperse polymers. Macromol. Symp. 113:177–196 (1997). J. M. H. M. Scheutjens, G. J. Fleer, and M. A. Cohen Stuart, End effects in polymer adsorption: a tale of tails. Colloids Surf. 21:285–306 (1986). A. N. Semenov, J. Bonet-Avalos, A. Johner, and J. F. Joanny, Adsorption of polymer solutions onto a flat surface. Macromolecules 29:2179–2196 (1996). M. A. Cohen Stuart, F. H. W. H. Waajen, T. Cosgrove, T. L. Crowley, and B. Vincent, Hydrodynamic thickness of adsorbed polymer layers. Macromolecules 17:1825–1830 (1984). C. W. Hoogendam, C. J. W. Peters, A. de Keizer, M. A. Cohen Stuart, and B. H. Bijsterbosch, Layer thickness of adsorbed CMC on hematite. J. Colloid Interface Sci. (submitted). M. A. Cohen Stuart, J. M. H. M. Scheutjens, and G. J. Fleer, Polydispersity effects and the interpretation of polymer adsorption isotherms. J. Poly. Sci. Poly. Phys. Ed. 18:559– 573 (1980). J. M. H. Scheutjens and G. J. Fleer, Some implications of recent polymer adsorption theory, in The Effect of Polymers on Dispersion Properties (Th. F. Tadros, ed.). Academic Press, London, 1982, pp. 145–168. S. P. F. M. Roefs, J. M. H. M. Scheutjens, and F. A. M. Leermakers, Adsorption theory for polydisperse polymers. Macromolecules 27:4810–4816 (1994). D. C. Leonhardt, H. E. Johnson, and S. Granick, Adsorption isotope effect for protioand deuteropolystyrene at a single surface. Macromolecules 23:685 (1990). M. A. Cohen Stuart, G. J. Fleer, and J. M. H. M. Scheutjens, Displacement of polymers. I: Theory. Segmental adsorption energy from polymer desorption in binary solvents. J. Colloid Interface Sci. 97:515–525 (1984). G. P. van der Beek, M. A. Cohen Stuart, G. J. Fleer, and J. E. Hofman, A chromatographic method for the determination of segmental adsorption energies of polymers. Polystyrene on silica. Langmuir 5:1180–1186 (1989). G. P. van der Beek, M. A. Cohen Stuart, G. J. Fleer, and J. E. Hofman, Segmental adsorption energies for polymers on silica and alumina. Macromolecules 24:6600–6611 (1991).


20

Cohen Stuart

25.

L. Vroman and A. L. Adams. Adsorption of proteins out of plasma and solutions in narrow spaces. J. Colloid Interface Sci. 111:391 (1986). E. Pefferkorn and A. Elaissari, Adsorption–desorption processes in charged polymer– colloid systems. Structural relaxation of adsorbed macromolecules. J. Colloid Interface Sci. 138:187 (1990). M. C. P. van Eijk and M. A. Cohen Stuart, Polymer adsorption kinetics: effects of supply rate. Langmuir 13:5447 (1997). J. Buijs, P. A. W. van den Berg, J. W. Th. Lichtenbelt, W. Norde, and J. Lyklema, Adsorption dynamics of IgG and its F(ab⬘)2 and Fc fragments studied by reflectometry. J. Colloid Interface Sci. 178:594–605 (1996); M. C. L. Maste, Proteolytic stability in colloid systems. Ph.D. thesis, Wageningen University, 1996. J. C. Dijt, M. A. Cohen Stuart, and G. J. Fleer, Surface exchange kinetics of chemically different polymers. Macromolecules 27:3229–3237 (1994). H. E. Johnson and S. Granick, Exchange kinetics between the adsorbed state and free solution: PMMA in CCl4. Macromolecules 23:3367–3374 (1990). H. E. Johnson, J. F. Douglas, and S. Granick, Regimes of polymer adsorption desorption kinetics. Phys Rev. Lett. 70:3267 (1990). A. Krabi and M. A. Cohen Stuart, Sequential adsorption of polymer—displacement or trapping? Macromolecules (1990). I. Borukhov, D. Andelman, and H. Orland, Polymer solutions between charged surfaces, Europhys. Lett. 32:499 (1995). J. C. Leyte and M. Mandel. Potentiometric behavior of poly(methacrylic acid). J. Pol. Sci. A2:1879–1891 (1964). J. Lyklema, Fundamentals of Interface and Colloid Science, Vol. II. Academic Press, London, 1995, Chap. 3. M. R. Bo¨hmer, O. A. Evers, and J. M. H. M. Scheutjens, Weak polyelectolytes between two surfaces: adsorption and stabilization. Macromolecules 23:2288–2301 (1990). H. G. M. van de Steeg, M. A. Cohen Stuart, A. de Keizer, and B. H. Bijsterbosch, Polyelectrolyte adsorption: a subtle balance of forces. Langmuir 8:2538–2546 (1992). M. A. Cohen Stuart, Polyelectrolytes on solid surfaces, in Short and Long Chains at Interfaces, Proc. 30th Rencontres de Moriond (J. Daillant, P. Guenoun, C. Marques, P. Muller, and J. Traˆn Thanh Vaˆn, eds.). Editions Frontie`res, 1995, pp. 3–12. N. G. Hoogeveen, M. A. Cohen Stuart, and G. J. Fleer, Polyelectrolyte adsorption on oxides. I: Kinetics and adsorbed amounts. J. Colloid Interface Sci. 182:133–145 (1996). C. W. Hoogendam, A. de Keizer, M. A. Cohen Stuart, B. H. Bijsterbosch, and J. G. Batelaan, Adsorption mechanics of carboxymethylcelluloses on mineral surfaces. Langmuir (submitted). M. A. Cohen Stuart, C. W. Hoogendam, and A. de Keizer, Kinetics of polyelectrolyte adsorption. J. Phys. Condensed Matter (1997). J. F. Joanny, Adsorption of a polyampholyte chain. J. Phys. France II 4:1281–1288 (1994). S. Neyret, L. Ouali, F. Candau, and E. Pefferkorn, Adsorption of polyampholytes on PS latex. Effect on the colloid stability. J. Colloid Interface Sci. 176:86–94 (1995). B. C. Bonekamp, Adsorption of polylysins at solid–liquid interfaces. Ph.D. thesis, Wageningen University, 1984.

26.

27. 28.

29. 30. 31. 32. 33. 34. 35. 36. 37. 38.

39. 40.

41. 42. 43. 44.


2 Driving Forces for Protein Adsorption at Solid Surfaces WILLEM NORDE The Netherlands

I.

Wageningen Agricultural University, Wageningen,

INTRODUCTION

Proteins are copolymers of some 22 different amino acids of varying hydrophobicity (Fig. 1). As a consequence, proteins are more or less amphiphilic and, therefore, usually highly surface active. Moreover, a number of amino acid residues in the side groups along the polypeptide chain contain positive or negative charges. This makes the protein an amphoteric polyelectrolyte. Based on the spatial organization in protein molecules the following distinctions may be made: 1.

2.

3.

Molecules that are highly solvated and flexible, resulting in a disordered (randomly) coiled structure. This group comprises some proteins of which the natural function is nutritional, such as the caseins in milk and glutelins in wheat grains. Furthermore, unfolding of ordered proteins, for instance by heat treatment or by adding a denaturant, often leads to a loosely coiled structure. Molecules that have adopted a regular structure (e.g., helices and pleated sheets), the so-called fibrillar proteins. Fibrillar proteins are usually found in connective tissue and they are often insoluble in water. Molecules that contain different structural elements, i.e., helices, pleated sheets, and parts that are unordered, which are folded up into a compact, densely packed structure: the globular proteins. By way of example, computer graphic images of the globular protein bovine pancreas ribonuclease are shown in Fig. 2. Although the globular proteins represent only a minority of the available protein mass, they make up by far the greatest proportion of protein species. These proteins have evolved to fulfill specific functions. This applies to enzymes, transport proteins, and immunoproteins (antibodies). The biological functioning of a given protein is directly related to its specific three-dimensional structure.

With respect to practical applications and implications the adsorption of the globular proteins is most relevant. Examples can be found in biomedical engineering, biosensors, immunological test systems, immobilized-enzyme bioreactors, biofouling of © 2003 by Marcel Dekker, Inc.

22

Norde

FIG. 1 Schematic representation of two peptide units in a polypeptide chain. Two of the three backbone bonds in the peptide unit are free to rotate; the other one is fixed. The amino acid side groups R, R⬘, . . . may be polar or apolar and may contain ionized groups.

processing equipment, various forms of chromatography, and many others. Therefore, this chapter focuses on the adsorption of globular proteins. The theoretical understanding of the adsorption of (randomly) coiled flexible polymers, including polyelectrolytes, has greatly advanced over the last few decades. However, because of their intricate highly specific structures, a general theory for globular protein adsorption has not been developed yet. At best some general principles may be indicated. In the next section, the theoretical trends, which as a rule are experimentally verified, for the adsorption of uncharged and charged flexible polymers are briefly summarized. A more detailed discussion on this subject is given by Martinus A. Cohen Stuart in Chapter 1. These trends may serve as a starting point in the discussion of globular protein adsorption in subsequent sections.

II.

TRENDS IN THE ADSORPTION BEHAVIOR OF UNCHARGED AND CHARGED FLEXIBLE POLYMERS

Flexible polymers in solution possess a high conformational entropy resulting from the various states each of the many segments in the polymer chain can adopt. The expansion of a polymer coil is determined by the quality of the solvent. The better the solvent, the more expanded the coil is and the higher its conformational entropy. Adsorption leads to a reduction of this conformational entropy. Hence, adsorption takes place only if the loss in conformational entropy is compensated by sufficient attraction between polymer segments and the surface. The critical Gibbs energy for adsorption to occur spontaneously is typically a few tenths of a kT unit per segment. Even if the Gibbs energy of adsorption is only slightly higher than the critical value, the whole polymer molecule adsorbs tenaciously and, apparently, irreversibly. This is because the contribution from each adsorbing segment adds to the Gibbs energy of adsorption of the whole polymer molecule. The resulting high affinity between the polymer and the surface is reflected in the shape of the adsorption isotherm (where the adsorbed mass ⌫ is plotted against the polymer concentration in solution after adsorption cp); the initial part of the isotherm merges with the ⌫-axis because © 2003 by Marcel Dekker, Inc.

Driving Forces for Protein Adsorption

23

FIG. 2 Computer graphic images of bovine pancreas ribonuclease, showing (top) the polypeptide backbone made up of ␣-helices (spirals seen for example in the lower right of the molecule), ␤-sheets (upper left) and ‘‘unordered’’ parts, and (bottom) a space-filling model showing the compact packing of a globular protein molecule.

at low polymer supply all of the polymer is adsorbed until the surface is saturated (Fig. 3). Figure 4 depicts how the segments of an adsorbed flexible polymer molecule may be distributed over trains, loops, and tails. Trains account for the attached segments; they are rarely very long and they do not completely cover the entire sorbent © 2003 by Marcel Dekker, Inc.

24

Norde

FIG. 3 High-affinity adsorption isotherm, typical for polymer adsorption.

surface, leaving about 20–30% of the surface uncovered. Loops account for most of the adsorbed mass. Their occurrence limits the reduction in conformational entropy. Their extension is determined by the solvent quality. A high loop density is tolerated close to the surface only if the solvent is relatively poor. For a poor solvent the adsorbed mass in the (pseudo-) plateau region of the isotherm typically is in the range of 2–4 mg m⫺2; for a good solvent it amounts to 0.5–1.0 mg m⫺2. For entropic reasons tails usually extend far in the solution and, therefore, play a dominant role in steric interaction. The experimental thickness of the adsorbed polymer layer depends on the method of determination. For instance, some methods, e.g., ellipsometry and reflectometry, determine an optical thickness (⬇ average loop extension), whereas other methods, such as dynamic light scattering and viscometry, yield the hydrodynamic thickness (⬇ tail extension). Invariably, the tails extend further from the sorbent surface than the average of the loops. Based on these considerations the adsorption behavior of flexible polyelectrolytes may be predicted. Because of the charge they carry, polyelectrolytes are strongly expanded in aqueous solution; in other words water is an excellent solvent for flexible polyelectrolytes. As a consequence the formation of loops is strongly suppressed. Hence, polyelectrolytes tend to adsorb in thin layers up to only a few tenths of a milligram per square meter. As with uncharged polymers, polyelectrolytes also require some critical attractive interaction with the sorbent to become adsorbed. In addition to an electrostatic contribution, the Gibbs energy of adsorption may also

FIG. 4 Conformations of a flexible polymer adsorbed on a surface: (a) poor solvent and (b) good solvent.



25

comprise a nonelectrostatic component. Depending on the charge signs of the polyelectrolyte and the surface, the electrostatic contribution is attractive or repulsive; it may or it may not outweigh the nonelectrostatic contribution. If too strongly repulsive, it prevents the polyelectrolyte from adsorbing. For sake of comparison, the electric part of the Gibbs energy of a monovalent ionic group in an electric field is ca. 1 kT for every 25 mV, and the Gibbs energy of dehydration of a — CH2 — group is ca. 1.5 kT, both at room temperature. Hence, polyelectrolytes with some hydrocarbon groups in their chain may readily adsorb on a hydrophobic surface against an unfavorable electric potential. In distinction to uncharged polymers, the adsorption of polyelectrolytes is very sensitive to indifferent low-molecular-weight electrolytes. These electrolytes exert a dual effect: (1) They screen the intramolecular repulsion between charged segments, which manifests itself in water becoming a poorer solvent. Therefore, the addition of salt promotes the formation of loops and hence results in more adsorbed mass per unit sorbent surface. (2) They also screen electrostatic interactions between a polymer segment and the sorbent surface. Attachment of the segment to the sorbent surface is promoted/opposed by electrolyte if this interaction is repulsive/attractive. Along similar lines the influence of the pH on the adsorption of polyelectrolytes containing weak ionic groups (e.g., carboxyl and/or amino groups) may be explained. At a pH where such a polyelectrolyte is uncharged it adsorbs in a relatively thick layer; the adsorbed amount is then high and independent of ionic strength. However, at a pH where the ionic groups are fully charged the adsorbed amount is low, but it increases with increasing salt concentration. For similar reasons amphiphilic polyelectrolytes show maximum adsorption at their isoelectric point. This maximum is less pronounced at higher salt concentrations.

III.

STRUCTURE AND STABILITY OF GLOBULAR PROTEINS IN AQUEOUS SOLUTION

Unlike the flexible polymers discussed in the foregoing section, globular proteins in aqueous solution acquire compact, ordered conformations in which the atomic packing density, expressed in volume fraction, reaches values of 0.70–0.80. In such a compact conformation the rotational freedom along the polypeptide chain is severely restricted, implying a low conformational entropy. The compact structure is possible only if interactions within the protein molecule and interactions between the protein molecule and its environment are sufficiently favorable to compensate for the low conformational entropy. These interactions are of different nature, i.e., Coulomb, hydrophobic, hydrogen bonding, and interactions between fixed and/or induced dipoles. Protein adsorption research is often focused on structural rearrangements in the protein molecules, not only because of its relevance to the biological functioning of the molecules, but also because of the significant role such rearrangements play in the mechanism of the adsorption process. Knowledge of the factors that determine the protein structure helps to understand the behavior of proteins at interfaces. The major factors will be briefly discussed. © 2003 by Marcel Dekker, Inc.

26

A.

Norde

Conformational Entropy of Proteins

Most globular proteins contain a significant amount (⬃40–80%) of ordered structural elements, notably ␣-helices and/or ␤-sheets. These structures are stabilized by hydrogen bonds between peptide units in the polypeptide backbone (Fig. 5). These intramolecular hydrogen bonds reduce the rotational mobility of the bonds in the polypeptide chain and, as a consequence, the conformational entropy. The dense packing of the polypeptide backbone and of the side groups within the interior of the folded molecule will further reduce the conformational entropy. Thus, for a protein molecule consisting of 100 amino acid residues (corresponding to a molar mass of ca. 10,000 Da) the conformational entropy decrease upon folding the polypeptide chain to a compact conformation containing ca. 50% ordered secondary structure (␣helix and/or ␤-sheet) is estimated to be several hundred joules per degree Kelvin, corresponding to a Gibbs energy increase of up to a few hundred kilojoules per mole at 300 K [1]. B.

Hydrophobic Interaction

Dehydration of nonpolar components in an aqueous environment results in an increase of the entropy of the water molecules released from those components and, therefore, in a lowering of the Gibbs energy of the system. This favorable hydrophobic dehydration causes apolar parts of the polypeptide in water to associate. The relevance of hydrophobic dehydration for protein folding was first recognized by Kauzmann [2], and it is now considered the primary driving force for the folding process. To estimate the contribution from hydrophobic interaction to the stabilization of a compact structure, the hydrophobicities of the constituting amino acids must be known. These hydrophobicities may be assessed by partitioning the amino acids between water and a nonpolar solvent. It has thus been established that the Gibbs energy of dehydration decreases by 9.2 kJ mol⫺1 after reducing the hydrophobic water-accessible surface area by 1 nm2 [3]. Then, assuming that ca. 60% of the protein’s interior consists of apolar amino acid residues [4], the contribution from hydrophobic dehydration to the Gibbs energy of stabilization of a compact globular molecule of 10,000 Da molar mass is about 500 kJ mol⫺1 at room temperature. C.

Coulomb Interaction

Most of the charged amino acid residues in a protein molecule are located in its aqueous periphery. Model calculations where the protein molecule is approximated by a sphere with its surface exhibiting evenly, but discretely, distributed positive and negative charges yields attractive Coulomb interaction in the isoelectric region and repulsive interaction remote from the isoelectric point [5–7]. Hence, near the isoelectric point Coulomb interaction is expected to favor a compact conformation, and at more extreme pH values a more expanded structure is promoted. It should, however, be realized that the Coulomb effects are largely dependent on the distribution of the charged residues on the protein molecule. For instance, ion pairs on the protein surface stabilize the compact conformation [8]. Furthermore, the electrolyte concentration largely affects the distance over which charged groups interact. © 2003 by Marcel Dekker, Inc.


27

FIG. 5 Ordering of polypeptide chains into an ␣-helix structure (left) and a parallel ␤-pleated sheet (right).

If charged residues are present in the interior of a globular protein molecule, they usually occur as ion pairs. Unfolding of the protein would imply rupture of the electrostatically favorable ion pairs but also leads to favorable hydration of both ionic groups. Because of these compensating effects, ion pairing in the apolar, nonaqueous interior of the protein does not make a major contribution to the stabilization of a compact protein structure taking the unfolded polypeptide chain as in the reference state [8]. Ionization of residues originally buried in the low-dielectric interior of the globular protein molecule in the nonionized form (e.g., histidine and tyrosine) may be a significant driving force for unfolding [9,10].

D.

Lifshitz–van der Waals Interactions

Lifshitz–van der Waals interactions originate from interactions between fixed and/ or induced dipoles. They are very sensitive to the separation distance r between the dipoles, varying as r⫺6. Upon folding the polypeptide chain into a compact structure, © 2003 by Marcel Dekker, Inc.

28

Norde

dipolar interactions between the protein and water are disrupted, but on the other hand dipolar interactions inside the protein molecule and between water molecules are newly formed. The overall effect of dipolar interactions on protein stability is not clear, although it is generally assumed that because of the relatively high packing density, dipolar interactions tend to promote a compact structure [11,12].

E.

Hydrogen Bonds

Most of the hydrogen bonds in globular proteins are those between amide and carbonyl groups of the polypeptide backbone. In ␣-helices and ␤-sheets such intrachain hydrogen bonds enforce each other because they are aligned more or less parallel to one another (Fig. 5). The number of hydrogen bonds involving amino acid side groups is, as a rule, relatively small. As with dipolar interactions, folding of the polypeptide chain implies loss of hydrogen bonds between protein and water and formation of intramolecular hydrogen bonds in the protein and among water molecules. The net effect on the stabilization of the protein structure remains unclear. Studies using model components [13–15] indicate that two peptide–water interactions on the one hand are more favorable than peptide–peptide and water–water interactions on the other. This would mean that hydrogen bonding does not stabilize a compact protein structure. However, if due to other types of interaction (e.g., hydrophobic bonding) the polypeptide chain and side chains that are capable of forming hydrogen bonds are forced in the apolar interior of a compact protein, the formation of intramolecular hydrogen bonds would largely stabilize that structure.

F.

Bond Lengths and Bond Angles

It is probable that in a tightly packed compact conformation not all lengths and angles of the covalent bonds attain the most favorable values. Indeed, energyminimization calculations point to distortion of covalent bonds in (crystallographic) globular proteins that significantly opposes the folded conformation, possibly up to several kilojoules per mole [1,16]. The various contributions that determine the protein structure in aqueous solution are summarized in Table 1. Hydrophobic interaction and conformational entropy of the protein are the major factors. Because of their opposing effects, which are of comparable magnitude, the folded globular conformation is thermodynamically only marginally stable. Typically, the Gibbs energy of stabilization of the globular structure (relative to the unfolded structure) is in the range of some tens of kilojoules per mole, corresponding to, e.g., the energy required to rupture a few hydrogen bonds in an apolar environment. Thus, although conformational entropy and hydrophobic dehydration dominate protein folding, none of the other factors are unimportant. As a consequence, (small) disturbances in the environment, e.g., changes in temperature, pH, ionic strength, etc., and the introduction of an interface, may induce perturbations in the protein structure. By way of example, Fig. 6 shows the Gibbs energy and the heat-induced unfolding of ␣-lactalbumin, together with its enthalpic and entropic components. It is © 2003 by Marcel Dekker, Inc.


29

TABLE 1 Contributions to the Gibbs Energy G of the Native (N) Globular Protein Structure Relative to the Unfolded Denatured (D) Structure

FIG. 6 Changes in the Gibbs energy, enthalpy, and entropy for the heat-induced denaturation of ␣-lactalbumin. Ionic strength is 0.05 M KCl.


30

Norde

clear that the small change in Gibbs energy is the result of substantial enthalpy– entropy compensation.

IV.

ADSORPTION OF PROTEINS FROM AQUEOUS SOLUTION ONTO SOLID SURFACES

Proteins are polymers. Their adsorption behavior shares some features with those for the flexible polymers, summarized in Section II and discussed more extensively in Chapter 1. First, an adsorbing protein molecule forms numerous contacts with the sorbent surface [15,17], usually leading to a high-affinity adsorption isotherm (Fig. 3). As a consequence, the rate of desorption upon diluting the system is often below the detection limit. However, adsorbed proteins may readily be exchanged against other surface-active solutes being either of the same or of another kind [18–22]. For instance, surface-active components of relatively low molecular weight may effectively strip off proteins from surfaces [23–25]. Furthermore, the sequential adsorption of different proteins from complex biological fluids is well documented [26,27]. Second, the adsorption pattern of proteins is typical for a polyampholyte; the amount adsorbed generally is at a maximum around the isoelectric point of the protein/ sorbent complex, i.e., at conditions where the charge on the protein and the sorbent just compensate each other. Also, similar to flexible polymers, proteins may tend to spread over the sorbent surface. However, due to their relatively strong internal cohesion the spreading rate of proteins is far slower than that of flexible polymers. If the rate of spreading is comparable to the rate of deposition at the surface, the extent of spreading (= conformation change) decreases with increasing flux toward the surface and, as a consequence, the adsorbed mass will be larger. Such behavior has been observed for proteins [28,29], whereas under usual experimental conditions the spreading of flexible polymers is much faster than the transport so that the adsorbed amounts are essentially independent of the flux [30]. Another, even more basic, difference in the adsorption behavior between globular proteins and flexible polymers concerns the change in the conformation entropy. In contrast to flexible polymers, globular proteins are highly ordered, low-entropy structures. Upon adsorption this structure may (partly) break down, thereby increasing the conformational entropy of the protein. Hence, where attachment to a surface always leads to a loss of conformational entropy of a flexible polymer [31], protein adsorption may result in increased conformational entropy. This entropy gain may be sufficiently large to cause spontaneous adsorption of the protein under otherwise adverse conditions, i.e., at a hydrophilic, electrostatically repelling surface [32,33]. In forthcoming sections the primary contributions to protein adsorption on a smooth, rigid surface will be discussed in more detail. (The influence of water-soluble oligomers or polymers preadsorbed or grafted on the sorbent surface will not be considered.) The major contributions originate from (1) redistribution of charged groups (ions) when the electrical double layers around the protein molecule and the sorbent surface overlap, (2) dispersion interaction between the protein and the sorbent, (3) changes in the hydration of the sorbent surface and of the protein molecule, and (4) structural rearrangements in the protein. Although these factors will be discussed more or less individually, it may be clear that their actions are interdependent, being either synergistic or antagonistic. For instance, the structural flexibility of an © 2003 by Marcel Dekker, Inc.


31

adsorbed protein molecule strongly affects the electrostatic and hydrophobic interaction between the protein and the sorbent.

A.

Redistribution of Charged Groups Resulting from Overlapping Electrical Double Layers

In an aqueous environment both the protein molecule and the sorbent surface are, as a rule, electrically charged. The charge on the protein arises from association or dissociation of acidic and basic amino acid residues. Depending on the type of sorbent material, its surface charge stems from the dissociation/association of (strong) acidic or basic groups (e.g., sulfonated, carboxylated, and aminated polymer latexes, oxide surfaces, and surfaces of biological materials) or from the binding of ions other than protons (e.g., silver halide surfaces). Other surfaces may not contain charged groups, such as those of polyethylene oxide, Teflon, etc. Both the protein and the sorbent are surrounded by counterions and co-ions. A fraction of these ions may be specifically absorbed to the protein and sorbent surface, whereas the remaining ions are diffusely distributed in the solution. The charge on the surface together with its compensating surrounding charge is referred to as the electrical double layer. Figure 7 gives a schematic representation of the electrical double layer according to the Gouy–Stern model [34]. The electrical part of the Gibbs energy of an electrical double layer Gel equals the isothermal, reversible work of charging it [35,36]:

冕

␴0

Ge1 =

␾0⬘ d␴0⬘

(1)

0

where ␾⬘0 and ␴⬘0 are the variable surface potential and surface charge density, respectively, during the charging process.

FIG. 7 Schematic representation of the Gouy–Stern model of an electrical double layer, indicating the distribution of the counterions (⫹) and co-ions (⫺), compensating the negative surface charge and the resulting potential decay.


32

Norde

Solving Eq. (1) requires knowledge of ␾0⬘(␴⬘). This functionality follows from 0 combining the Poisson equation [Eq. (2)] with an expression for the charge distribution in the electrical double layer. At any point in the electrical double layer the electrostatic potential ␾(x) is related to the volume charge density ␳(x) by ⵜ2␾(x) = ⫺

␳(x) ␧␧0

(2)

where ␧␧0 is the dielectric permittivity of the medium considered, and ␧0 that of free space. To evaluate ␾(x), assumptions have to be made concerning the distribution of charge across the electrical double layer. For instance, for charged surfaces the Gouy–Stern model (Fig. 7) is usually adopted, where the surface charge is located at x = 0 and where the countercharge is diffusely distributed in the solution, separated from the surface by a charge-free layer having a thickness equal to the radius of a hydrated counterion. Then, taking into account the proper boundary conditions, which for a planar surface are

冉冊 ⭸␾ ⭸x

x=r

=

␴x=r ␧0␧x=r

(3)

where x = r refers to any value of x in the electrical double layer, Eq. (2) can be solved and ␾0⬘(␴0⬘) can be evaluated. In turn, Gel can be computed using Eq. (1). When the protein and the surface approach each other, their electrical double layers overlap, giving rise to electrostatic interaction (Fig. 8). The resulting change in Gibbs energy, ⌬adsGel, is the difference between Gel after and before adsorption. It follows that Eq. (1) has to be applied three times, i.e., for the protein-covered sorbent surface, for the bare sorbent surface, and for the protein molecule in solution. According to this procedure, various attempts have been made to calculate ⌬adsGel. Sta˚hlberg and Jo¨nsson [37,38] thus compared the interaction for constant charge densities at the surfaces with that for constant potentials. Their data on protein retention in ion exchange chromatography could be reasonably well described by the constant-potential approach. It points to a charge regulation when the protein comes

FIG. 8 Schematic representation of a protein–sorbent surface system before and after adsorption. The charged groups on the surface and the protein molecule are indicated by ⫹/⫺, and the low-molecular-weight ions are indicated by 䊝/䊞. The shaded areas represent hydrophobic regions.



33

close to the stationary phase. A similar approach has been undertaken by Lenhoff and coworkers [39–41]. They emphasized the effect of the heterogeneous charge distribution in the protein molecule. Proteins having a relatively large dipole moment, such as ribonuclease [39] and chymotrypsinogen A [40,41], are expected to adsorb in a preferred orientation, whereas the less dipolar lysozyme [40,41] approaches the sorbent surface in a more random orientation. Similarly, Roush et al. [42] computed a preferred orientation of cytochrome b5 on an anion exchange surface. A more detailed electrostatic analysis was undertaken by Noinville et al. [43], who computed interaction energies for all atoms in a protein/sorbent system. In this way they predicted a preferred orientation of ␣-lactalbumin and lysozyme at a poly(vinyl imidazole) surface. The values obtained for ⌬adsGel in the simulation studies mentioned depend on parameters such as charge densities (or potentials) at the surfaces of the protein and the sorbent, the separation distance between these two components, the size and the orientation of the protein molecule in the electric field, the ionic strength, and the dielectric constant of the surrounding medium. Thus, in an aqueous environment of 0.1 M ionic strength and at a separation between the sorbent and the protein molecule of, say, 1 nm the value for ⌬adsGel typically varies from a few RT up to a few tens of RT per mol of protein. At close ‘‘atomic’’ contact between protein and sorbent the aforementioned approaches may not be satisfactory. In particular, they may not properly account for charge regulation. Experimental and theoretical analyses show that under many conditions the Debye length is smaller than the dimensions of the protein molecule. For instance, in a 0.1 M solution of a 1:1 electrolyte in water the Debye length is ca. 1 nm, whereas the thickness of an adsorbed protein layer usually is at least a few nanometers. In that case, the protein layer shields the sorbent–protein contact region from the solution. It implies that the sorbent surface charge should be largely compensated by countercharge in the protein–sorbent contact region. The extent of charge neutralization was predicted by Norde and Lyklema [44,45] based on a threelayer model for the adsorbed protein. This model, depicted in Fig. 9, assumes coverage of the sorbent surface by a compact protein layer (cf. Section IV.D). All sorbent surface charge is located at x = 0. The inner region 1, 0 < x ⱕ m, contains a fraction of the adsorbed protein charge and any ions trapped between the protein of the sorbent surface. The thickness of region 1 is of the order of the diameter of a hydrated ion, which is in the range of a few tenths of a nanometer. The extension of the outer region 3, p ⱕ x ⱕ d, is assumed to be comparable to the distance over which charged groups (including their hydration layer) on the protein surface protrude into the aqueous medium; for this distance 0.7 nm is taken. Analogous to the interiors of native state globular proteins, the central region 2, m < x < p, is considered to be void of isolated charged groups. The thickness of this region follows from measured thicknesses (see Section IV.D) of adsorbed protein layers corrected for the thicknesses assumed for regions 1 and 3. Because of the requirement of overall electroneutrality,

␴0 ⫹ ␴1 ⫹ ␴2 ⫹ ␴3 ⫹ ␴d = 0

(4)

where the indices refer to the sorbent surface, the three regions of the adsorbed layer, and the diffuse part of the electrical double layer, respectively. Based on the assumptions that ␴0 is located at x = 0, that ␴1, ␴2 (= 0), and ␴3 are distributed © 2003 by Marcel Dekker, Inc.

34

Norde

FIG. 9 Model for the adsorbed protein–sorbent interface, in which the course of the electrostatic potential is indicated. For details refer to the text.

homogeneously over regions 1, 2, and 3 and that ␴d is exponentially distributed according to the Gouy–Stern model (Fig. 7), Norde and Lyklema [45] derived expressions for ␾(x) across the adsorbed layer and within the aqueous solution. By applying Eq. (2) a qualitative representation of ␾(x) is shown in Fig. 9. For all possible values of ␴0 (derived from titration data for the bare sorbent surface [46]) and ␴d (derived from electrokinetic data [47]), ␾(x) shows a strong dependence on the assumed division of charge between regions 1 and 3. Since region 1 has a relatively low dielectric permittivity, any net charge in the contact zone between the protein and the sorbent leads to a high electrostatic potential and is therefore highly unfavorable. Realistically, it is not to be expected that ␾(x) attains values exceeding a few hundred millivolts. Hence, any mismatch of protein and sorbent charge in region 1 must be compensated by low-molecular-weight ions to make this region nearly electrically neutral. An example of charge compensation predicted by the model is represented by the curve in Fig. 10; it shows a strong dependence on the electrical states of the protein and the sorbent. Here, the number of counterions taken up was estimated from calculated values of ␴1, which follow from reasonable estimates for ␾m (e.g., ⫺100 mV) and the relation [45] d␾m p ⫺ m d⫺p = ⫹ d␴1 ␧0␧2 2␧0␧3

(5)

Equation (5) indicates that ␾m is highly sensitive to changes in ␴1 and, hence, to the number of coadsorbed ions. The tendency of charge neutralization is experimentally confirmed by the shift in the proton titration curves of proteins upon adsorption at a charged surface [48– 51]. By way of example, proton titration curves of ␣-lactalbumin in solution and © 2003 by Marcel Dekker, Inc.


35

FIG. 10 Charge compensation in desorbed layers of human serum albumin on a negatively charged polystyrene surface (␴0 = ⫺15.5 ␮C cm⫺2) as predicted from the model (——). Experimentally determined incorporation of cations in the adsorbed layer from 0.02 M BaCl2 solution (⫻) or 0.02 M MnCl2 solution (䊱) at 25⬚C.

adsorbed on both positively and negatively charged polystyrene latexes are presented in Fig. 11. Charge adjustments may also occur on the sorbent surface. This has been clearly demonstrated by Fraaije for the adsorption of bovine serum albumin on silver iodide crystals [20]. Apart from adjustments on the protein and, possibly, the sorbent surface, the charge density in the protein–sorbent contact region may be further regulated by the transfer of indifferent electrolyte between that region and the solution. Indications for such transfer were inferred from electrokinetic data [44,49], and direct experimental evidence was obtained by tracing radiolabeled ions [52]. Experimental data on the uptake of ions are included in Fig. 10. It is clear that the number of coadsorbed cations tends to increase with increasing pH, i.e., with increasing charge antagonism between the protein and the sorbent surface. Moreover,

FIG. 11 Proton titration curve for ␣-lactalbumin in solution (——) and adsorbed on positively (⭈ ⭈ ⭈ ⭈) and negatively (– – – –) charged polystyrene surfaces. Ionic strength 0.05 M; 25⬚C.


36

Norde

on the basis of phenomenological (thermodynamic) linkage relations the partitioning of ions between the adsorbed protein layer and the solution can be derived from the electrolyte dependency of the protein adsorption isotherms [53]. So we conclude that by active participation of small ions in the overall adsorption process, adverse conditions of charge on the protein and the sorbent surface may be eliminated. As a consequence of this charge regulation the absolute value of ⌬adsGel does not exceed a few tens of RT, and it is not very sensitive to the charge on the protein (or, for that matter, the pH) and on the sorbent surface. Its sign and value primarily depend on the charge distributions and the dielectric constants of the electrical double layers before and after the adsorption process, respectively [45,54]. In addition to charge regulation, the transfer of ions from an aqueous to a nonaqueous protein layer includes a chemical effect as well. Because the proteinaceous environment is a poorer ‘‘solvent’’ for the low-molecular-weight ions compared to water, the chemical effect opposes protein adsorption [34,54]. Hence, if a protein is repelled from a like-charged surface, it is due to the unfavorable chemical effect that would result from the incorporation of ions needed to regulate the charge in the protein–sorbent contact region. Indeed, maximum protein adsorption affinity has been found under conditions where the charge on the protein just compensates the charge on the sorbent surface so that no additional ions are needed [20]. Finally, it should be realized that all these approaches, except the one by Noinville et al. [43], are based on continuum models. Each of the components, i.e., the sorbent surface, the protein molecule, and the solvent, are treated as continuous media having their own, constant dielectric permittivity, and in each of the phases the charge is smeared out according to the assumed distribution. However, in reality, in particular at high ionic strength, the distances between the individual charged groups on the protein and on the sorbent surface may exceed the Debye length, so a discrete-charge model would be more appropriate. Furthermore, as a result of the polar heterogeneity of most protein molecules, different areas in the adsorbed layer may have different permittivities. Including these specific effects in the analysis would not only require exact knowledge of the three-dimensional protein structure, but also detailed information concerning the orientation and structural perturbation of the protein upon adsorption. For most systems such detailed information is not available. B.

Dispersion Interaction

The most advanced computation of dispersion interaction between macroscopic bodies is based on quantum electrodynamics of continuous media, as presented in the Lifshitz theory [55]. A more approximate, simpler treatment is that in which the interaction energy is obtained by pairwise summation of London–van der Waals energies between all molecules of the interacting bodies. This approach has been elaborated by Hamaker [56] and de Boer [57]. The Hamaker–de Boer approximation may deviate by 10–30% with respect to the absolute magnitude [34]. However, since the (adsorbed) protein molecules and, sometimes, the sorbent material are often not sufficiently well defined to allow for quantum-dynamic computations, it is reasonable to utilize the Hamaker–de Boer theory for these systems. Then, for a sphere interacting with a planar surface the contribution from van der Waals interaction to the Gibbs energy of adsorption, ⌬adsGvdW, is given by © 2003 by Marcel Dekker, Inc.


⌬adsGvdW = ⫺

A132 6

冉

a a h ⫹ ⫹ ln h h ⫹ 2a h ⫹ 2a

37

冊

(6)

where A132 is the Hamaker constant for the interaction between body 1 (the sorbent) and body 2 (the protein) across a medium 3 (the aqueous solution) and where a is the radius of the sphere and h the distance of closest approach between the sphere and the planar surface. For h A3 and A2 > A3, so A132 > 0 and hence ⌬adsGvdW < 0, which implies attraction. The Hamaker constant for interaction across water is ca. 6.6 ⫻ 10⫺21 J for proteins [11], 1–3 ⫻ 10⫺19 J for metals [34] and 4–12 ⫻ 10⫺21 J for synthetic polymers, e.g., polystyrene and Teflon [34]. According to Eq. (6a), ⌬adsGvdW increases linearly with increasing dimensions of the protein molecule and it drops off steeply (hyperbollically) with increasing separation distance between the protein and the sorbent. For example, for a spherical protein molecule of radius 3 nm at a distance of 0.1 nm from the sorbent surface, ⌬adsGvdW at room temperature is calculated to be 1–4 RT/mol in case of a synthetic polymer surface and 6–11 RT at metal surfaces. The values given here may not be better than a qualitative indication for the magnitude of van der Waals interaction in protein adsorption. Obtaining more accurate values is hampered by the irregular geometry of the protein molecule and, possibly, the sorbent surface. Furthermore, structural perturbation in the adsorbed protein may affect the Hamaker constant to an unknown extent. C.

Hydration Changes

When the surfaces of the protein molecule and the sorbent are hydrophilic, their hydration is favorable. Therefore, if protein adsorption occurs it is likely that some hydration water is retained between the sorbent surface and the adsorbed layer. However, when (one of) the contacting surfaces are (is) hydrophobic, dehydration would simulate the adsorption process. 1.

Protein Hydrophobicity

Protein molecules contain both polar and apolar parts. As discussed in Section III.B, in globular proteins in an aqueous environment the apolar residues tend to be buried in the interior of the molecule where they are shielded from contact with water. However, due to other interactions that are participating in determining the protein structure and to geometrical constraints, as a rule not all apolar parts are hidden in © 2003 by Marcel Dekker, Inc.

38

Norde

the interior and not all the polar parts are exposed at the aqueous periphery of the protein molecule. For relatively small proteins (e.g., lysozyme, ␣-lactalbumin, ribonuclease, cytochrome c, etc.) having a molar mass of ca. 15,000 Da, apolar atoms occupy about 40–50% of the water-accessible surface area [4]. For larger proteins, which have a smaller surface/volume ratio the apolar fraction of the surface is usually less. Furthermore, water-soluble, nonaggregating proteins show a more or less even distribution of the polar and apolar residues over their surface so that no pronounced hydrophobic patches are present. Several studies [31,61,62] confirm that the hydrophobicity of the protein influences its adsorption. However, not only the hydrophobicity of the protein surface, but the overall hydrophobicity of the protein may be relevant for the adsorption behavior. The overall hydrophobicity influences the extent of structural perturbation upon adsorption, which, in turn, affects the adsorption affinity (cf. Section IV.D). As a consequence, it is virtually impossible to establish the influence of protein hydrophobicity, as such, on the adsorption process. 2.

Sorbent Hydrophobicity

As for the protein, the impact on adsorption by the sorbent hydrophobicity is difficult to establish experimentally because a variation in the hydrophobicity usually involves a variation in the electrostatic potential (or charge) as well. Experiments using hydrophobicity gradients [63] are probably best for using this matter in more detail. The contribution from dehydration of a component to the Gibbs energy of adsorption ⌬adsGhydr may be estimated from partition coefficients of (model) compounds in water/nonaqueous two-phase systems (e.g., Ref. 64). It has thus been established that, at room temperature, dehydration of most hydrophobic surfaces is predominantly entropically determined, involving effects in the range of 20–50 ␮J K⫺1 m⫺2; this corresponds to a lowering of the Gibbs energy with 5–12 mJ m⫺2. For a protein of molar mass 15,000 Da that adsorbs ca. 1 mg m⫺2, it corresponds to ⌬adsGhydr ranging between ⫺30 and ⫺75 RT/mol of adsorbed protein. It demonstrates that hydrophobic dehydration easily overrules effects from the redistribution of charged groups and from dispersion interaction (cf. Sections IV.A and IV.B). D.

Rearrangements in the Protein Structure

As discussed in Section III, the three-dimensional structure of a globular protein molecule in aqueous solution is only marginally stable, and interaction with an interface may induce structural perturbations in the protein. However, unlike flexible polymers globular proteins do not adsorb in a loosely structured train loop tail–like conformation. The general observation that the thickness of an adsorbed protein layer is comparable with the dimensions of the native molecule in solution [65–68] points to a compact structure of the adsorbed protein molecules, even if they have undergone structural rearrangements (Fig. 8). In Section IV.A it has been explained that the formation of such a compact layer requires coadsorption of low-molecular-weight ions in the contact region between the protein and the sorbent to prevent a high electrostatic potential in that region. An alternative way to avoid the development of a high potential would be the unfolding of the protein into an open and loose structure that is freely penetrable for water and electrolyte, as depicted in Fig. 4. In such © 2003 by Marcel Dekker, Inc.


39

a layer the dielectric permittivity would not differ too much from that of the bulk solution. Because of the general observation that globular proteins do not form such loose adsorbed structures, it is concluded that the unfavorable chemical effect of ion incorporation (Section IV.A) is less unfavorable than the exposure of apolar residues of the protein to water, as would occur upon extensive unfolding. When a protein molecule arrives at the sorbent surface, at one side of the molecule the aqueous environment is replaced by the sorbent material. As a consequence, intramolecular hydrophobic interaction becomes less important as a structure-stabilizing factor; i.e., apolar parts that tend to be buried in the interior of the dissolved molecule may become exposed to the sorbent surface without making contact with water. Because hydrophobic interaction between amino acid side groups in the protein’s interior support the formation of secondary structures as ␣-helices and ␤-sheets, a reduction of this interaction tends to destabilize such structures. Decrease in ␣-helix and/or ␤-sheet content is indeed expected to occur only if peptide units released from these structures can form hydrogen bonds with the sorbent surface, as is the case for oxides (e.g., glass, silica, and metal oxides) or with residual water molecules remaining at the sorbent surface. Then the decrease in secondary structure may lead to an increased conformational entropy of the protein, contributing to ⌬adsG with a few tens of RT per mol of protein [23,69–71] (cf. Section III.A). If in the nonaqueous protein–sorbent contact region it is not possible for the peptide units to form hydrogen bonds with the sorbent surface, as is the case for hydrophobic surfaces, adsorption may induce formation of extra intramolecular peptide–peptide hydrogen bonds, thereby promoting the formation of ␣-helices and/or ␤-sheets [71,72]. Hence, whether adsorption on a hydrophobic surface results in an increased or decreased order in protein structure depends on the subtle balance between energetically favorable interactions and the protein’s conformational entropy.

V.

CONCLUSIONS

Protein adsorption is a complex process that is controlled by a number of subprocesses, the major ones being (1) electrostatic interactions between the protein and the sorbent surface, giving rise to coadsorption of small ions, (2) dispersion interaction, (3) changes in the state of hydration of the sorbent surface and parts of the protein molecule, and (4) structural rearrangements in the protein. These subprocesses are not independent of each other. For instance, hydrophobic bonding between the protein and the surface requires close contact between the two components, which may be optimized by structural rearrangements in the protein. This may involve decreased or increased flexibility along the polypeptide chain. At hydrophilic surfaces, increased conformational freedom is anticipated and indeed experimentally found, which in turn improves the ability of the protein to form ion pairs with oppositely charged groups on the sorbent surface. The synergistic and antagonistic effects of these interactions are indicated in Scheme 1. Based on these considerations it is to be expected that all proteins adsorb on hydrophobic surfaces, even under electrostatically adverse conditions. With respect to their adsorption behavior on hydrophilic surfaces, distinction may be made between structurally stable (‘‘hard’’) and labile (‘‘soft’’) proteins. The hard proteins adsorb on hydrophilic surfaces only if they are electrostatically attracted. The soft proteins are more prone to undergo © 2003 by Marcel Dekker, Inc.

40

Norde

SCHEME 1 Interdependency of the major subprocesses that are involved in the overall protein adsorption process. Adsorption promotion is denoted by ⫹ and adsorption opposition by ⫺.

structural rearrangements when they adsorb, and the ensuing gain in conformational entropy may be sufficiently large to cause adsorption on a hydrophilic electrostatically repelling surface.

REFERENCES 1. 2. 3. 4. 5.

T. E. Creighton, Proteins : Structures and Molecular Properties, 2nd ed. W. H. Freeman, New York, 1993, Chap. 5. W. Kauzmann, Some factors in the interpretation of protein denaturation. Adv. Protein Chem. 14:1–63 (1959). E. M. Richards, Areas, volumes, packing and protein structure. Ann. Rev. Biophysics. Bioeng. 6:151–176 (1977). B. Lee and F. M. Richards, The interpretation of protein structures: estimation of static accessibility. J. Mol. Biol. 55:379–400 (1971). C. Tanford and J. G. Kirkwood, Theory of protein titration curves. I. General equations for impenetrable spheres. J. Am. Chem. Soc. 79:5333–5339 (1957).


Driving Forces for Protein Adsorption 6. 7. 8. 9. 10. 11. 12.

13. 14. 15. 16. 17. 18. 19. 20.

21.

22. 23. 24. 25.

26. 27.

28. 29.

41

C. Tanford and J. G. Kirkwood, Theory of protein titration curves. II. Calculations for simple models at low ionic strength. J. Am. Chem. Soc. 79:5340–5347 (1957). J. B. Matthew and F. R. N. Gurd, Calculation of electrostatic interactions in proteins. Methods Enzymol. 130:413–453 (1986). D. J. Barlow and J. M. Thornton, Ion pairs in proteins. J. Mol. Biol. 168:867–885 (1983). F. Franks and D. Eagland, The role of solvent interactions in protein conformation. Crit. Rev. Biochem. 3:165–219 (1975). J. A. Schelman, Solvent denaturation. Biopolymers 17:1305–1322 (1978). S. Nir, van der Waals interactions between surfaces of biological interest. Prog. Surf. Sci. 8:1–58 (1977). C. A. Haynes, K. Tamura, H. R. Ko¨rfer, H. W. Blanch, and J. M. Prausnitz, Thermodynamic properties of aqueous ␣-chymotrypsin solutions from membrane osmometry measurements. J. Phys. Chem. 96:905–912 (1991). G. C. Kresheck and I. M. Klotz, The thermodynamics of transfer of amides from an apolar to an aqueous solution. Biochemistry 8:8–11 (1969). C. Tanford, Protein denaturation. Part C. Theoretical models for the mechanism of denaturation. Adv. Protein Chem. 24:1–95 (1970). W. Norde, Adsorption of proteins from solution at the solid–liquid interface. Adv. Colloid Interface Sci. 25:267–340 (1986). M. Levitt, A simplified representation of protein conformations for rapid simulation of protein folding. J. Mol. Biol. 104:59–107 (1976). B. W. Morrissey and R. R. Stromberg, The conformation of adsorbed blood proteins by infrared bound fraction measurements. J. Colloid Interface Sci. 46:152–164 (1974). J. L. Brash and Q. M. Samak, Dynamics of interactions between human albumin and polyethylene surface. J. Colloid Interface Sci. 65:495–504 (1978). B. M. C. Chan and J. L. Brash, Conformational change in fibrinogen desorbed from glass surface. J. Colloid Interface Sci. 84:263–265 (1981). J. G. E. M. Fraaije, Interfacial thermodynamics and electrochemistry of protein partitioning in two-phase systems. Ph.D. thesis, Wageningen Agricultural University, The Netherlands, 1987. V. Ball, A. Bentaleb, J. Hemmerle, J.-C. Voegel, and P. Schaff, Dynamic aspects of protein adsorption onto titanium surfaces: mechanism of desorption into buffer and release in the presence of proteins in the bulk. Langmuir 12:1614–1621 (1996). P. R. Van Tassel, P. Viot, and G. Tarjus, A kinetic model of partially reversible protein adsorption. J. Chem. Phys. 106:761–770 (1997). W. Norde and J. P. Favier, Structure of adsorbed and desorbed proteins. Colloids Surf. 64:87–93 (1992). W. Norde and A. C. I. Anusiem, Adsorption, desorption and re-adsorption of proteins on solid surfaces. Colloids Surf. 66:299–306 (1992). M. C. Wahlgren and T. Arnebrant, Interaction of cetyltrimethylammonium bromide and sodium dodecyl sulfate with ␤-lactoglobulin and lysozyme at solid surfaces. J. Colloid Interface Sci. 142:503–511 (1991). L. Vroman and A. L. Adams, Adsorption of proteins out of pasma and solutions in narrow spaces. J. Colloid Interface Sci. 111:391–402 (1986). T. A. Horbett, Adsorption of proteins from plasma to a series of hydrophilic– hydrophobic copolymers. II. Compositional analysis with the prelabeled technique. J. Biomed. Mater. Res. 15:673–695 (1981). J. J. Ramsden, Concentration of protein deposition kinetics, Phys. Rev. Lett. 71:295– 298 (1993). M. C. P. van Eijk and M. A. Cohen Stuart, Polymer adsorption kinetics: effects of supply rate. Langmuir 13:5447–5450 (1997).


42

Norde

30.

J. C. Dijt, M. A. Cohen Stuart, and G. J. Fleer, Reflectometry as a tool for adsorption studies. Adv. Colloid Interface Sci. 50:79–101 (1994). J. M. H. M. Scheutjens and G. J. Fleer, Statistical theory of the adsorption of interacting chain molecules. 1. Partition function, segment density distribution, and adsorption isotherms. J. Phys. Chem. 83:1619–1635 (1979). T. Arai and W. Norde, The behavior of some model proteins at solid–liquid interfaces. 1. Adsorption from single protein solutions. Colloids Surf. 57:1–16 (1990). C. A. Haynes and W. Norde, Globular proteins at solid–liquid interfaces. Colloids Surf. B: Biointerfaces 2:517–566 (1994). J. Lyklema, Fundamentals of Interface and Colloid Science. Volume I: Fundamentals. Academic Press, London, 1991. E. J. W. Verwey and J. T. G. Overbeek, The Theory of the Stability of Lyophobic Colloids. Elsevier, Amsterdam, 1948. A. J. Babchin, Y. Gur, and I. J. Lin, Repulsive interface forces in overlapping electric double layers in electrolyte solutions. Adv. Colloid Interface Sci. 9:105–141 (1978). J. Sta˚hlberg, U. Appelgren, and B. Jo¨nsson, Electrostatic interactions between a charged sphere and a charged planar surface in an electrolyte solution. J. Colloid Interface Sci. 176:397–407 (1995). J. Sta˚hlberg and B. Jo¨nsson, Influence of charge regulation in electrostatic interaction chromatography of proteins. Anal. Chem. 68:1536–1544 (1996). B. J. Yoon and A. M. Lenhoff, Computation of the electrostatic interaction energy between a protein and a charged surface. J. Phys. Chem. 96:3130–3134 (1992). C. M. Roth and A. M. Lenhoff, Electrostatic and van der Waals contributions to protein adsorption: computation of equilibrium constants. Langmuir 9:962–972 (1993). C. M. Roth and A. M. Lenhoff, Electrostatic and van der Waals contributions to protein adsorption: comparison of theory and experiment. Langmuir 11:3500–3509 (1995). D. J. Roush, D. S. Gill, and R. C. Willson, Electrostatic potentials and electrostatic interaction energies of rat cytochrome b5 and a simulated anion-exchange adsorbent surface. Biophys. J. 66:1290–1300 (1994). V. Noinville, C. Vidal-Madjar, and B. Se´bille, Modeling of protein adsorption on polymer surfaces. Computation of adsorption potential. J. Phys. Chem. 99:1516–1522 (1995). W. Norde and J. Lyklema, The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces. IV. The charge distribution in the adsorbed state. J. Colloid Interface Sci. 66:285–294 (1978). W. Norde and J. Lyklema, Thermodynamics of protein adsorption. Theory with special reference to the adsorption of human plasma albumin and bovine pancreas ribonuclease at polystyrene surfaces. J. Colloid Interface Sci. 71:350–366 (1979). K. Furusawa, W. Norde, and J. Lyklema, A method for preparing surfactant-free polystyrene latices of high surface charge. Kolloid-Z Z Polymere 250:908–909 (1972). W. Norde and J. Lyklema, The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces. III. Electrophoresis. J. Colloid Interface Sci. 66:277–284 (1978). W. Norde and J. Lyklema, The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces. II. Hydrogen ion titrations. J. Colloid Interface Sci. 66:266–276 (1978). C. A. Haynes, E. Sliwinski, and W. Norde, Structural and electrostatic properties of globular proteins at a polystyrene–water interface. J. Colloid Interface Sci. 164:394– 409 (1994). F. Galisteo and W. Norde, Protein adsorption at the AgI–water interface. J. Colloid Interface Sci. 172:502–509 (1995). F. Galisteo and W. Norde, Adsorption of lysozyme and ␣-lactalbumin on poly(styrene sulphonate) latices. II. Proton titrations. Colloids Surf. B: Biointerfaces 4:389–400 (1995).

31.

32. 33. 34. 35. 36. 37.

38. 39. 40. 41. 42.

43.

44.

45.

46. 47.

48.

49.

50. 51.


Driving Forces for Protein Adsorption 52. 53.

54. 55. 56. 57.

58. 59.

60. 61. 62.

63.

64. 65.

66.

67. 68.

69.

70. 71. 72.

43

P. van Dulm, W. Norde, and J. Lyklema, Ion participation in protein adsorption at solid surfaces. J. Colloid Interface Sci. 82:77–82 (1981). J. G. E. M. Fraaije, R. M. Murris, W. Norde, and J. Lyklema, Interfacial thermodynamics of protein adsorption, ion co-adsorption and ion binding in solution. I. Phenomenological linkage relations for ion exchange in lysozyme chromatography and titration in solution. Biophys. Chem. 40:303–315 (1991). W. Norde and J. Lyklema, Why proteins prefer interfaces. J. Biomater. Sci. Polymer Edn. 2:183–202 (1991). L. D. Landau and E. M. Lifshitz, Electrodynamics of Continuous Media, 2nd ed. Pergamon, Oxford, 1963. H. C. Hamaker, The London–van der Waals attraction between spherical particles. Physica. 4:1058–1072 (1937). J. H. de Boer, Influence of van der Waals forces and primary bonds in binding energy, strength, and orientation, with special reference to some artificial resins. Trans. Faraday Soc. 32:10–38 (1936). J. Visser, Hamaker constants. Comparison between Hamaker constants and Lifshitz–van der Waals constants. Adv. Colloid Interface Sci. 3:331–363 (1972). T. Afshar-Rad, A. I. Baily, P. F. Luckham, W. MacNaughtan, and D. Chapman, Forces between proteins and model polypeptides adsorbed on mica surfaces. Biochem. Biophys. Acta. 915:101–111 (1987). J. N. Israelachvili, Intermolecular and Surface Forces, 2nd ed. Academic Press, New York, 1992. F. E. Regnier, The role of protein structure in chromatographic behavior. Science 238: 319–323 (1987). A. A. Gorbunov, A. Y. Lukyanov, V. A. Pasechnik, and A. V. Vakrushev, Computer simulation of protein adsorption and chromatography. J. Chromatog. 365:205–212 (1986). H. Elwing, A. Askendal, U. Nilsson, and I. Lundstro¨m, A wettability gradient method for studies of macromolecular interactions at the liquid–solid interface. J. Colloid Interface Sci. 91:248–255 (1987). G. Ne´methy, H. A. Scheraga, Structure of water and hydrophobic bonding in proteins. J. Chem. Phys. 36:3401–3417 (1962). N. de Baillou, P. Dejardin, A. Schmitt, and J. L. Brash, Fibrinogen dimensions at an interface: variations with bulk concentration, temperature and pH. J. Colloid Interface Sci. 100:167–174 (1984). P. Schaaf and P. Dejardin, Structural changes within an adsorbed fibrinogen layer during the adsorption process: a study by scanning angle reflectometry. Colloids Surf. 31:89– 103 (1988). E. Blomberg, P. M. Claesson, and R. D. Tilton, Short-range interaction between adsorbed layers of human serum albumin. J. Colloid Interface Sci. 166:427–436 (1994). E. Blomberg, P. M. Claesson, J. C. Fro¨berg, and R. D. Tilton, Interaction between adsorbed layers of lysozyme studied with the surface force technique. Langmuir 10: 2325–2334 (1994). A. Kondo, F. Mukarami, and K. Higashitani, Circular dichroism studies on conformational changes in protein molecules upon adsorption on ultrafine polystyrene particles. Biotechnol. Bioeng. 40:889–894 (1992). A. Kondo, S. Oku, and K. Higashitani, Structural changes in protein molecules adsorbed on ultrafine silica particles. J. Colloid Interface Sci. 143:214–221 (1991). T. Zoungrana, G. H. Findenegg, and W. Norde, Structure, stability, and activity of adsorbed enzymes. J. Colloid Interface Sci. 190:437–448 (1997). M. C. L. Maste, W. Norde, and A. J. W. G. Visser, Adsorption-induced conformational changes in the serine proteinase savinase: a tryptophan fluorescence and circular dichroism study. J. Colloid Interface Sci. 196:224–230 (1997).


3 Thermodynamics of Adsorption of Amino Acids, Small Peptides, and Nucleic Acid Components on Silica Adsorbents VLADIMIR A. BASIUK Instituto de Ciencias Nucleares, Universidad Nacional Autońoma de Me´xico, Mexico City, Mexico

I.

INTRODUCTION

Adsorption of amino acids, peptides, and nucleic acid constituents on silica surfaces is a very important aspect of biomolecular adsorption. One cannot imagine the modern practice of their analysis and separation, so widely used in biotechnology, medicine, diagnostics, etc., without high-performance liquid chromatographic (HPLC) methods employing silica stationary phases (see, for example, Refs. 1–3). The rational design of new biocompatible materials and drug delivery systems also needs a detailed knowledge on biomolecule adsorption interactions. Furthermore, studying the adsorption of the biopolymers’ building blocks and the development of structure– property relationships can give insight to the interfacial behavior of the biopolymers themselves. A key point in the characterization of adsorption processes on solid–liquid interfaces is determination of the corresponding thermodynamic parameters: Gibbs free energy (⌬G ⬚), enthalpy (⌬H ⬚), and entropy (⌬S⬚) changes during adsorption. Now adsorption from solutions on solid surfaces is a highly advanced theory accounting for different possible types of adsorption isotherms and allowing a detailed quantitative description of various particular systems to be performed [4–8]. In practice, however, ‘‘a reasonable balance between theoretical foundations and derivations, on the one hand, and results, on the other hand’’ is necessary [5]. In this regard, thermodynamics of adsorption of amino acids, small peptides, and nucleic acid constituents on silica surfaces is especially illustrative because, as is demonstrated in this chapter, so far it has been described using only the simplest experimental approaches and the simplest numeric interrelations. In particular, the case of adsorption was always considered from very diluted aqueous solutions; this fact can be readily explained by the dominating role of the works on HPLC separations in the whole body of publications related to biomolecular adsorption. Due to the same circumstances, particular types of adsorption isotherm and real surface nonuniformity are also neglected. © 2003 by Marcel Dekker, Inc.

46

Basiuk

II.

EXPERIMENTAL APPROACHES TO DETERMINE ADSORPTION THERMODYNAMIC CONSTANTS

A.

Static Approach

The traditional approach to the determination of adsorption thermodynamic constants is based on measuring adsorption isotherms under static conditions. The adsorption equilibrium constant (or distribution coefficient) K can be relatively easily found from a slope of the isotherm initial rectilinear part; this then enables calculation of the free energy change during adsorption ⌬G⬚: ⌬G⬚ = ⫺RT ln K

(1)

Measuring the isotherms under several different temperatures and plotting ln K versus 1/T (Van’t Hoff plots) enables us to derive the entropy and enthalpy changes ⌬S⬚ and ⌬H⬚. One should note, however, that for solutes which are adsorbed more weakly, the results will be less accurate. Moreover, measuring the adsorption isotherms themselves is not always possible. This is the case when adsorption is very slight, e.g., the equilibrium constants are close to one, especially if K < 1. The latter case means that molecule concentration in the solid phase appears less than in the bulk liquid phase (negative excess adsorption), and it is impossible to detect any concentration changes in the solution. B.

Dynamic Approach

The dynamic (liquid chromatographic, usually HPLC) approach is based on measuring a compound’s retention value k⬘ on a chromatographic column [9–11]. The equilibrium constant in this case is equated to K=

k⬘ ␾

(2)

where ␾ is the column phase ratio and k⬘ is the capacity factor for a given solute, k⬘ =

(ti ⫺ t0) (Vi ⫺ V0) = t0 V0

(3)

where ti and Vi are retention time and retention volume of the solute i; t0 and V0 are retention time (usually called the dead time) and retention volume (the dead, or void, volume equivalent to the total volume of eluent in the column) for a compound which is not retained (or not adsorbed). The simplest definition for ␾ is the ratio of volumes of stationary (Vs) to mobile (Vm) phases in the chromatographic column [10]. A more strict definition of the phase ratio accounts for the mass (m) and specific surface area S of the stationary phase [12]:

␾=

mS Vs␳S = Vm Vm

(4)

where ␳ is the density of the stationary phase. The free energy change during adsorption ⌬G ⬚ is then calculated according to Eq. (1). Performing the chromatographic procedure under different temperatures allows us to derive the corresponding ⌬S⬚ and ⌬H⬚ values from Van’t Hoff plots. © 2003 by Marcel Dekker, Inc.

Thermodynamics of Adsorption

47

Like the static approach, this method has its drawbacks as well. The most serious problem is that in practice a column dead volume V0 (or t0) and phase ratio are relatively difficult to determine precisely [11,13]. An additional problem appears when solutes chromatographed possess relatively low capacity factors k⬘, for instance, when k⬘ < 1 precise measurement is difficult; as is shown subsequently, this is the case for most amino acids. Thus it seems quite reasonable to consider some thermodynamic parameters derived from HPLC experiments as ‘‘questionable’’ [11]. Realizing these problems, it is preferable to discuss not determination, but only estimation of these parameters. Nevertheless, when the same HPLC column (and the same chromatographic system as a whole) is employed for a series of the measurements, errors due to the dead volume and phase ratio are the same for all solutes studied. As a result, any graphic interrelation between thermodynamic parameters and solute molecular properties, e.g., linear trends for the plots ⌬G ⬚ versus number of aliphatic carbon atoms found for amino acids and peptides (see Sections III.A–C), are maintained but are shifted along the ordinate, e.g., the ⌬G ⬚ axis, depending on these errors. Even if the latter might appear significant, one can discuss comparative adsorption behavior of solutes. For instance, regardless of the errors due to incorrect measurements of the dead volume and phase ratio, a solute retained for a longer time in a chromatographic column will have a higher ⫺⌬G⬚ (or lower ⌬G ⬚) value. Suppose two compounds A and B are chromatographed on the same column and have the capacity factors k⬘A and k⬘B, respectively. Their adsorption equilibrium constants, KA and KB, depend on the same ␾ value, which can be completely excluded if we consider not corresponding ⌬G⬚A and ⌬G⬚B by themselves, but their difference ␦(⌬G⬚AB), i.e., a relative free energy change:

␦(⌬G⬚AB) = ⌬G⬚A ⫺ ⌬G⬚B = ⫺RT ln

KA k⬘A = ⫺RT ln = ⫺RT ln ␣A/B KB k⬘B

(5)

Equation (5) includes one more parameter, ␣A/B, which is the selectivity of resolution of A and B. Conventionally, the compound retained longer is designed as A, i.e., k⬘A > k⬘, B in order to obtain ␣A/B > 1. Such description is indispensable in the HPLC resolution of enantiomeric compounds to characterize their chiral discrimination (see, for example, Ref. 14). Values of ⌬S⬚ are also affected by the errors due to incorrect measurements of the dead volume and phase ratio, whereas ⌬H ⬚ values found from slopes of the Van’t Hoff plots are not and thus appear to be the most reliable thermodynamic parameters derived from HPLC retention data.

III.

THERMODYNAMICS OF ADSORPTION OF SMALL BIOMOLECULES

A.

Amino Acids

Despite obvious implications of amino acid adsorption on silica adsorbents for their liquid chromatographic analysis, and continuously growing number of publications on the HPLC of amino acids, there are not many works focused on the thermodynamic aspects. Nevertheless, to get a general idea on their comparative behavior, even barely chromatographic data can be useful. A good example is the paper by © 2003 by Marcel Dekker, Inc.

48

Basiuk

Molna´r and Horva´th [15]. The authors found that under isocratic elution (i.e., the solvent composition is maintained the same during the chromatographic procedure) of a homologous series of ␣-amino acids having n-alkyl side chains on a column with octadecyl silica, the relationship between the log k⬘ values and the carbon number of the side chain is linear (Fig 1). The capacity factors of the species with undissociated carboxylic groups (at pH 0.2) are higher than those of the about half dissociated species (at pH 2.1), indicating that on octadecyl-silica surface the former are adsorbed stronger than the latter. Admitting that both chromatographic retention on nonpolar stationary phases in aqueous media and partitioning between organic solvents and aqueous solutions have a similar physicochemical basis, the authors compared the partitioning coefficients, P, for protein amino acids in octanol–water and the retention values obtained for octyl silica (pH 6.7) and found another linear relationship, shown in Fig. 2. An attempt to establish a direct relationship between the structure of amino acids and the free energy changes has been undertaken for the case of adsorption on bare silica from purely aqueous solutions using the HPLC approach [16–18]. The choice of pure water as the adsorption medium excluded any interference of the

FIG. 1 Plots of log k⬘ against the side chain carbon number for the homologous series of ␣-amino acids. Stationary phase, LiChrosorb RP-18 octadecyl silica; eluents: 0.5 M HClO4, pH 0.2, and 0.1 M phosphate buffer, pH 2.1; temperature 70⬚C. (From Ref. 15.)



49

FIG. 2 Plot of log P versus log k⬘ for the series of ␣-amino acids. Stationary phase, LiChrosorb RP-8 octyl silica; eluent: 0.1 M phosphate buffer, pH 6.7, temperature 70⬚C. (From Ref. 15.)

components of buffer solutions and simplified extremely the adsorption system. As it was found from the experimental retention values using Eqs. (1) and (2), for the overwhelming majority of ␣-amino acids the values of ⌬G ⬚ were positive (ranging from 130 J mol⫺1 for Phe to 3640 J mol⫺1 for Asp; for Pro, ⫺400 J mol⫺1) and K < 1, indicating that the molecular concentration in the solid phase is lower than in the bulk liquid phase. As was mentioned in Section II.A, the traditional static approach cannot be successful for thermodynamic characterization of such adsorption systems. Plotting the ⌬G ⬚ values versus the number of aliphatic carbon atoms in the molecules revealed a linear interrelation for n-alkyl (aliphatic bifunctional) ␣-amino acids, similar to that reported by Molna´r and Horva´th [15]; this is shown in Fig. 3. From the slope, an increment in ⌬G ⬚ for each aliphatic C atom has been obtained to be about ⫺300 J mol⫺1. In addition, amino acid adsorbability is considerably influenced by heteroatoms and other nonaliphatic moieties in the ␣-substituent. Imidazole nucleus (for His) and carboxylic groups (Asp and Glu) cause the sharpest increase in ⌬G ⬚ values (and corresponding decrease in adsorbability); amide (Asn and Gln) and alcohol functions (Ser and Thr) also reduce adsorbability but to a much lesser extent. The sulfur atom noticeably decreases ⌬G ⬚ in the case of Cys (as compared to Ala) but only slightly contributes in the case of Met (as compared to Val). The ⌬G⬚ decrease can also be due to phenyl nucleus (Phe versus Ala), but © 2003 by Marcel Dekker, Inc.

50

Basiuk

FIG. 3 Plot of ⌬G ⬚ versus the number of aliphatic carbon atoms (nC) in the amino acid molecules for the adsorption on bare silica from water. Aliphatic bifunctional amino acids and proline (●) and other amino acids (䉭). Temperature 19⬚C. (From Ref. 18.)

only if the nucleus does not contain hydroxy substituents, as in Tyr and 3,4-dioxyphenylalanine (DOPA). Proline is adsorbed much stronger than its open-chain analog valine, which is apparently explained by a cyclic structure of Pro. We attempted a comparison with some other molecular and thermodynamic properties of amino acids [17]. For example, the relationship between ⌬G⬚ and apparent molar volume showed a good fit to a straight line. Plotting the heats and free energies of amino acid formation versus ⌬G⬚ also revealed trends close to linear, but the fits were much worse. Some idea of the accuracy of the dynamic determination of K and ⌬G ⬚ values has been provided from the following estimates [17]. The error of recording the retention volumes typically did not exceed 2%. Taking a solute with the average retention volume Vi of 200 ␮L as an example, due to this error it can vary from 196 to 204 ␮L. Correspondingly, the under- and overestimation of the K values gives 0.70 and 0.76 (4% deviation from the average value of 0.73 calculated for Vi of 200 ␮L); for the free energy changes, the values of 870 and 670 J mol⫺1 are obtained (12 to 14% deviation from the average ⌬G ⬚ value of 760 J mol⫺1), respectively. The problem of precise determination of the dead volume and phase ratio still remained an open issue. A more careful approach is to avoid any estimates of the free energy changes and focus on the enthalpy change determination from Van’t Hoff plots. Uddin et al. [19] did this to study Gln, Met, Phe, and Trp adsorption on bare silica surface from water at 20, 37, and 55⬚C. They used a liquid chromatographic technique as well, though it was not HPLC. Nevertheless, evident advantage of their experimental setup was the use of the method of disturbance peaks, perhaps the most precise method to determine the column dead volume [13]. Thus, the K values presented should be highly reliable and should give rise to reliable corresponding ⌬G⬚ values. In Table 1 the latter are presented for 20⬚C along with the enthalpy changes derived from the Van’t Hoff plots (Fig. 4a). As shown, both the ⌬G ⬚ and ⌬H ⬚ values increase in the © 2003 by Marcel Dekker, Inc.


51

TABLE 1 Comparison of the Free Energy (for 20⬚C) and Enthalpy Changes for the Adsorption of Phe, Met, Trp, and Gln on Bare Silica from Water Obtained from Dynamic Measurements by Two Different Groups ⌬G 2⬚0 (J mol⫺1) Amino acid Gln Met Phe Trp

⌬H ⬚ (J mol⫺1)

From Ref. 19a

From Ref. 20

From Ref. 19

From Ref. 20

460 ⫺830 ⫺1530 ⫺490

1040 760 280 910

⫺2470 ⫺3190 ⫺6030 ⫺2910

⫺2150 ⫺9450 ⫺11620 ⫺8110

a

Calculated from the reported K values.

same order of Gln > Trp > Met > Phe. This, as well as the fact that the K values decrease (Fig. 4a) and ⌬G ⬚ values correspondingly increase with temperature, indicates that the adsorption process is enthalpy driven. It is interesting to compare these results with our recently published data for the same adsorption systems, also obtained using the dynamic (but HPLC) approach [20]. An important difference in the experimental setup was that we used a microcolumn of 64 ⫻ 2 mm I.D., whereas Uddin et al. [19] used a bigger column, 100 ⫻ 11 mm I.D. Figure 4 demonstrates that the two sets of the Van’t Hoff plots are very similar though shifted along the ln K axis. While in Fig. 4a the plots are situated basically above the abscissa, our series appears completely below it. Consequently, ⌬G⬚ values in the latter case are all positive. Considerable discrepancies are found also for the ⌬H ⬚ values (Table 1). One should note, however, that both the ⌬G⬚ and ⌬H⬚ values increase in the same order, Gln > Trp > Met > Phe, as in the work by Uddin et al. Revalidation of all the quantitative results would by highly desirable. In addition to Gln, Met, Phe, and Trp, we performed the estimates of ⌬H⬚ for many other ␣-amino acids as well as attempted to determine complementary ⌬S⬚

FIG. 4 Van’t Hoff plots for the adsorption of Phe, Met, Trp, and Gln on bare silica from water. [(a) From Ref. 19. (b) From Ref. 20.]


52

Basiuk

values [20]. As a whole, a significant scatter was observed for the points in the Van’t Hoff plots (Fig. 4b exemplifies this). In most cases, the correlation coefficients [20] exceeded 0.75, reaching 0.97 (Ala and Phe). However, for Ser and Asn their values appear unsatisfactory: 0.31 and 0.38, respectively. Therefore the estimates of ⌬H⬚ and ⌬S⬚ for these two amino acids cannot be considered reliable at all. The ⌬G ⬚ values increase with temperature in all cases due to enthalpy-driven adsorption. All the ⌬H⬚ and ⌬S⬚ values are negative. The ⌬H ⬚ values vary from ⫺550 J mol⫺1 for Ser to ⫺6030 J mol⫺1 for Phe; the ⌬S⬚ values, from ⫺6.6 J mol⫺1 K⫺1 for Asn to ⫺24.0 J mol⫺1 K⫺1 for Asp. The latter data suggest the adsorption process to be unfavorable. An interesting and useful way to analyze the thermodynamic data derived from the HPLC measurements is the approach based on enthalpy–entropy compensation. This phenomenon manifests itself in a linear interrelation between the free energy changes and the corresponding enthalpy changes for intrinsically similar classes of solutes [21,22]. If such interrelation is observed, the related equilibrium process is said to be isoequilibrium one. In other words, compounds which exhibit compensation behavior are considered to be adsorbed according to an essentially identical mechanism. Conventionally, compensation behavior is verified by linearity of the plots of ⌬H ⬚ versus ⌬S⬚ [21,22]. However, under such conditions the linearity sometimes arises not only from the compensation itself, but also from statistical effects due to possible errors associated with the determination of enthalpy changes [23,24]. On the other hand, if the plots of ⌬H ⬚ versus ⌬G⬚ are used, the statistical effects are minimized and the linearity is indicative of real compensation behavior. We also analyzed our data from this point of view, plotting ⌬H ⬚ versus ⌬S⬚ and ⌬G⬚ [20]. For both interrelations, the points did not completely fit onto straight lines (Figs. 5 and 6), apparently due to experimental errors. Nevertheless, the trends toward linearity were evident. The points for amino dicarboxylic acids, Glu and Asp, are separate from those for the majority of amino acids; this was clearly seen, es-

FIG. 5 Plot of ⌬H ⬚ versus ⌬S ⬚ for the adsorption of amino acids on bare silica from water. (From Ref. 20.)



53

FIG. 6 Plot of ⌬H ⬚ versus ⌬G ⬚ (at 20⬚C) for the adsorption of amino acids on bare silica from water. (From Ref. 20.)

pecially for the plot of ⌬H⬚ versus ⌬G⬚20 (Fig. 6). This may be due to the presence of a second carboxylic group in the molecules, resulting in certain differences between a mechanism of Glu and Asp adsorption and that for other amino acids. Some suggestions on the nature of these differences can be made based on the available data on ion equilibria in amino acid aqueous solutions. Accounting for the dual nature of amino acid molecules, such equilibria can be discussed in terms of acid and base pKi values as well as corresponding isoelectric points (pI ). Use of the latter is more appropriate in the present case, since the pI values reflect overall charge of the molecules. It is known that Glu and Asp have isoelectric points at 3.22 and 2.77, respectively; whereas pI values for other amino acids mentioned in Figs. 5 and 6 vary from 5.41 (Asn) to 6.30 (Pro) [25]. Thus, none of them exceeded the almost neutral pH values of the diluted aqueous solutions. Under such conditions, an equilibrium exists between the zwitterionic form and the form having a deprotonated ⫺ amino group, i.e., NH⫹ and NH2CHRCOO⫺, respectively. For Glu and 3 CHRCOO Asp, which possess much lower isoelectric points, the concentration of the zwitterions is lower, and the concentration of negatively charged species is correspondingly higher than for other amino acids, resulting in the observed deviation from the common linear trend. To elucidate whether the adsorption equilibrium correlates with the ionic equilibria in the solutions, we tested the graphic interrelation between our experimental ln K values and amino acid isoelectric points. As a result, a general linear regularity was obtained (Fig. 7), similar to the plots for ⌬H⬚ versus ⌬S⬚ and ⌬G⬚. Lower ln K values correspond to higher pI values, i.e., amino acids having higher isoelectric points are, as a rule, adsorbed on bare silica more strongly. This is evidence for major contribution of electrostatic interactions to the adsorption process and against hydrophobic interactions, since no direct correlation between our equilibrium data and the amino acid hydrophobicity indices has been found [20]. © 2003 by Marcel Dekker, Inc.

54

Basiuk

FIG. 7 Interrelation between ln K (at 20⬚C; bare silica) and isoelectric points of amino acids, pI. (From Ref. 20.)

The latter fact seems quite natural for the case of bare silica; its surface contains ionizable — — Si — OH groups and no hydrophobic hydrocarbon moieties. Nevertheless, the complementary adsorption studies on a hydrophobic octadecyl silica, widely used in HPLC, demonstrated that the close-to-linear interrelation between the equilibrium constants and pI is observed in this case as well [20]. What could well be expected here is a linear correlation between the equilibrium data and the amino acid hydrophobicity indices because amino acids are retained on C18 phases according basically to a hydrophobic mechanism. To verify this, we plotted the hydrophobicity coefficients (16 sets compiled and developed by Wilce et al. [26]) versus the ln K values. In most cases the plotting gave rather random distribution of points; however, in a few cases linear trends were found (Fig. 8). Another important difference, as compared to the adsorption on bare silica, was that at ambient temperature, for example, hydrophobic amino acids (Val, Nva, Leu, Ile, Nle, Tyr, Phe, Trp, and Met) have K > 1, i.e., the solute concentration in the adsorbed phase is higher than that in the bulk liquid phase (positive excess adsorption). The ⌬G ⬚ values cover the range of ⫺8400 (Trp) to 3220 J mol⫺1 (Glu). Similarly to the case of bare silica, K values decrease and ⌬G⬚ values increase with temperature. Statistical treatment of the Van’t Hoff plots revealed a very good fit of the points to straight lines for all amino acids; the correlation coefficients varied from 0.97 (Phe and Trp) to 1.00 (Asp and Met) [20]. Thus, the reliability of the thermodynamic estimates in the case of octadecyl silica is much higher than in the case of bare silica. The ⌬H ⬚ values were found to range from ⫺6310 J mol⫺1 for Asp to ⫺31360 J mol⫺1 for Trp, being negative for all amino acids, indicating enthalpy-driven adsorption. The ⌬S ⬚ values, as a whole, were much more negative compared to those for adsorption on bare silica—the lowest value was derived for Trp, ⫺79.7 J mol⫺1 K⫺1, and Asp had the highest one, ⫺32.2 J mol⫺1 K⫺1. The latter data suggest that the adsorption process is unfavorable, as in the case of bare silica. Analyzing the thermodynamic data for possible compensation behavior, we found that points in the plot of ⌬H ⬚ versus ⌬G ⬚ fit to a straight line better than those for amino acid ad© 2003 by Marcel Dekker, Inc.


55

FIG. 8 Interrelations between the experimental ln K values (at 20⬚C) for the adsorption of amino acids on octadecyl silica from water and some hydrophobicity coefficients. Hydrophobicity increases as hydrophobicity coefficient increases for FAUC, MEEK, MEEKR, and GUO, or as hydrophobicity coefficient decreases for HW. For abbreviations and references, see Ref. 26. (From Ref. 20.)

sorption on bare silica. Regarding the amino dicarboxylic acids Glu and Asp, the corresponding points were separate from the general regularity in the plot ⌬H⬚ versus ⌬S⬚, whereas they fit the linear trend of ⌬H ⬚ versus ⌬G ⬚. Thus we concluded that the adsorption behavior of amino acids on octadecyl silica is more uniform than that on bare silica, i.e., all the amino acids studied are adsorbed according to an essentially identical mechanism. All the preceding thermodynamic data refer to the amino acid adsorption on bare and octadecyl silicas from pure water and by no means are directly applicable to other solvent systems. Both pH variations and even minor additives of other compounds (and all the more components of conventional buffers) can drastically change the adsorption characteristics. This provides almost unlimited opportunities for optimizing the chromatographic resolution of amino acids and related compounds by shifting the ion equilibria, introducing ion-pairing reagents, surfactants, etc. (see, for example, Ref. 27). One of the most interesting HPLC developments afforded in this way was the resolution of amino acid enantiomers on conventional C18 silicas dynamically modified with substituted amino acids containing a long (C7 –C18) linear hydrocarbon radical (e.g., [28]). Such compounds are strongly adsorbed and not washed away by aqueous solvents. After complexation of the adsorbed molecules with copper(II) ions, the resulting stationary phase can be used for an efficient resolution of amino acid racemates. In our studies [29], one of the stationary phases used for this purpose was C18 silica dynamically modified with N-octyl-L-proline. Eluting racemic amino © 2003 by Marcel Dekker, Inc.

56

Basiuk

acids (Val, Nva, Leu, Nle, Phe, Tyr, and Met) with water containing 10⫺4 –10⫺3 M CuSO4, we were able to afford their enantioseparation with the selectivity parameter ␣D/L [Eq. (5)] ranging from 1.62 (Met, 10⫺3 M CuSO4) to 2.51 (Leu, 10⫺4 M CuSO4). According to the same Eq. (5), it is easy to calculate the corresponding relative free energy changes ␦(⌬G⬚D/L): ca. ⫺1170 to ⫺2240 J mol⫺1, respectively. A plethora of similar thermodynamic estimates can be derived from the published selectivity values; however, they are considered of secondary importance in HPLC resolution of enantiomers due to merely practical orientation of such works. On the contrary, the chromatographic approach to the determination of relative free energies of interaction between hydrophobic and amphiphilic amino acid side chains, applied by Pochapsky and Gopen [30] to a series of amino acid N-acetyl C-(N⬘-methyl)amides, can appear useful just from a fundmental point of view, e.g., for studying the thermodynamics of protein folding. As one can see, a lot of information on the thermodynamics of amino acid adsorption on silicas can be obtained by using the dynamic method. And what is the situation with the traditional batch technique? I am aware of no publications reporting on the amino acid adsorption thermodynamics from batch-measured isotherms. The isotherms alone, however, have been presented for several amino acids and modified silicas, which enables rough estimates of the free energy changes. In particular, Kubota et al. [31] studied the adsorption of glycine, leucine, histidine, and lysine on silica chemically modified with zirconium phosphate and aminobenzenesulfonic groups: Si

(CH 2) 3

N

N

SO 3H

The authors focused on the maximum adsorption capacities for amino acids and did not pay special attention to the initial parts of the isotherms, so that now it is difficult to determine their slopes more or less precisely. Based on the available data, one can say that for the silica modified with aminobenzenesulfonic groups (Fig. 9) the ⌬G⬚ values increase in the series Gly < His < Leu < Lys (adsorbability from the diluted solution increases in the reverse order), whereas the maximum adsorption capacity increases in the order Leu < Lys < His < Gly; i.e., there is no direct correlation between the two parameters. For the adsorption on the silica modified with zirconium phosphate groups (Fig. 10), the orders of increasing the maximum adsorption capacity and adsorbability from the diluted solution do not coincide as well, His < Lys < Gly = Leu and Lys < His < Gly < Leu, respectively. On average, the ⌬G⬚ values vary in the range of ⫺14.2 to ⫺9.1 kJ mol⫺1. B.

Linear Peptides

Logically, peptides should be adsorbed stronger than the amino acids constituting them: the longer the peptide chain, the stronger should be the adsorption. Such a cumulative effect ultimately results in a high adsorbability of proteins. For the simplest case of linear dipeptides adsorbed on bare silica from water, the free energy changes have been estimated from HPLC retention data [17,18]. Most dipeptides, similar to amino acids, have positive ⌬G⬚ values and K < 1; the exceptions are L-Val-L-Val and Gly-L-Leu (Fig. 11). For the dipeptides derived solely from © 2003 by Marcel Dekker, Inc.


57

FIG. 9 Adsorption isotherms of the amino acids on silica modified with aminobenzenesulfonic groups and the ⌬G ⬚ values derived from the isotherm initial parts. Temperature 25⬚C; pH 4. (From Ref. 31.)

aliphatic bifunctional amino acids, the dependence of ⌬G⬚ on the number of aliphatic C atoms does not fit well onto a straight line (Fig. 11, thick line). However, if we classify these dipeptides into those derived from Gly (Gly-Gly, Gly-DL-Ala, Gly-LVal, and Gly-L-Leu) and homodipeptides (again Gly-Gly, Ala-Ala diasteromers, and L-Val-L-Val), two separate linear trends can be obtained (Fig. 11, lower and upper fine lines, respectively). The increments in ⌬G ⬚ for each aliphatic C atom are about ⫺390 and ⫺260 J mol⫺1, respectively, i.e., adsorbability increases more sharply in the first series of dipeptides. A linear interrelation was also found between the ⌬G ⬚ values and apparent molar volumes for the dipeptides Gly-Gly, Gly-DL-Ala, and GlyL-Leu [17]. Similar to the case of amino acids (Section III.A), the presence of heteroatoms in the ␣-substituent substantially influences the adsorption characteristic of dipeptides [17,18]. The highest ⌬G⬚ value of 1170 J mol⫺1 was found for L-His-L-Leu, containing imidazole ring. The amide grouping of Asn moiety also decreases adsorbability, whereas indolyl and phenyl nuclei increase it. In the series of glycine dipeptides (Gly-DL-Ala, Gly-DL-Asn, and Gly-DL-Phe), the change of Ala to Asn residue increases ⌬G ⬚ by 320 J mol⫺1; the change of Ala to Phe lowers it by 720 J mol⫺1. A qualitatively similar picture can be found for the alanine dipeptides (DL-Ala-DLAla, DL-Ala-DL-Asn and DL-Ala-DL-Trp). An important aspect of the comparative characterization of amino acid and peptide adsorption behavior is the influence of peptide chain length. From compar© 2003 by Marcel Dekker, Inc.

58

Basiuk

FIG. 10 Adsorption isotherms of the amino acids on silica modified with zirconium phosphate groups and the ⌬G ⬚ values derived from the isotherm initial parts. Temperature 25⬚C; pH 4. (From Ref. 31.)

FIG. 11 Plot of ⌬G ⬚ versus the number of aliphatic carbon atoms (nC) in the linear dipeptide molecules for the adsorption on bare silica from water. Dipeptides derived from solely aliphatic bifunctional (●) and other amino acids (䉭). Lines, see text. Temperature 19⬚C. (From Ref. 18.)



59

ison of the data for amino acids and related peptides, it is clearly seen that the latter have better adsorbability (Figs. 3 and 11). In particular, the difference in the ⌬G⬚ values for Gly and Gly-Gly is 230 J mol⫺1; for DL-Ala and DL-Ala-DL-Ala, 510 J mol⫺1; and for L-Val and L-Val-L-Val, 890 J mol⫺1. These values are the increments in ⌬G ⬚ for each amino acid residue, and they increase as the number nC in amino acid residue increases. If we attribute the increments to the corresponding nC values, we obtain approximately the same values for the cited three pairs of solutes: 230, 255, and 220 J mol⫺1 (this result is rather close to that for the series of aliphatic bifunctional amino acids). Of special importance (e.g., for understanding the behavior of higher-molecular-weight peptides and proteins) is the question of how the adsorption characteristics change upon lengthening the peptide chain. For example, when studying the factors influencing retention of peptides in HPLC, Meek and Rosetti [32] found no general effect of the chain length. However, the authors considered a large number (100) of peptides with a random sequence and amino acid composition. For a valid comparison closely related compounds should be chosen, as was done by Molna´r and Horva´th [15]. Considering the series of L-Ala to its hexamer chromatographed at pH 0.2 and 2.1 on a C18 silica, they plotted log k⬘ versus the number of Ala residues to yield straight lines (Fig. 12). The slopes of the straight lines were much smaller than the slopes obtained with aliphatic bifunctional amino acids (Fig. 1) under identical chromatographic conditions. The log k⬘ increment of a methylene group of the aliphatic

FIG. 12 Plots of log k⬘ versus the number of residues in alanine oligomers. Stationary phase, LiChrosorb RP-18 octadecyl-silica; eluents: 0.5 M HClO4, pH 0.2, and 0.1 M phosphate buffer, pH 2.1; temperature 70⬚C. (From Ref. 15.)


60

Basiuk

side chain was roughly twice as large as the log k⬘ increment of the Ala residue. For phenylalanine and its oligomers, a similar linear dependence of log k⬘ on the number of Phe residues can be obtained for up to (at least) Phe5 [33]. A direct linear interrelation between the free energy changes and the number of amino acid residues was found for the adsorption of glycine and its oligomers (up to Gly4) on bare silica from water (Fig. 13a) [17,18]. The slope gives the ⌬G⬚ increment of ⫺220 J mol⫺1 per one Gly residue. Even Gly4 appears still to have a positive ⌬G⬚ value (700 J mol⫺1); extrapolation indicates that only the homopeptides beginning with Gly8 will have negative ⌬G⬚ values (positive excess adsorption) in this system. How long can such linear trends be maintained upon lengthening the peptide chain? For the case of reversed-phase HPLC, several research groups reported the linear relationships for peptides larger than eight amino acid residues (see Ref. [34] and references therein). Substantially longer peptides do not obey this regularity. This is generally assumed to be due to stabilization of secondary and tertiary structures in the oligo- and polypeptides, which remove some amino acid residues from contact with the silica surface. Obviously, the peptide chain folding should depend strongly both on the chemistry of silica surface and solvent composition, and no general predictions can be made so far. Even the simplest systems with the shortest peptides can exhibit deviations from the linear interrelation. This can best be exemplified by the retention of the same glycine series on the same bare silica considered in Fig. 13a, but using mixed acetate buffer–acetonitrile solvents instead of pure water [18]. As is seen from Figure 13b, a linear trend is no longer found for the plot of log k⬘ (and consequently ⌬G ⬚) versus nGly. Short peptide chain folding/unfolding in chromatographic systems is a relatively fast process compared to that for long peptides and proteins. While the former are always detected as a single peak, slow kinetics of long peptide and protein unfolding upon adsorption on the stationary phase results in split peaks corresponding to two stable confirmations [35,36]; one is more stable in the solution, whereas the other is more stable on the surface. The single compound thus can be characterized by two different ⌬G ⬚ values; nevertheless, in the present context only the value

FIG. 13 (a) Plot of ⌬G ⬚ versus the number of glycine residues (nGly) for the adsorption on bare silica from water. Temperature 19⬚C. (b) Plot of log k⬘ versus nGly. Mobile phase: acetate buffer solution (pH 5.21)–acetonitrile with volume ratio of (1) 45:55 and (2) 25:75. Temperature 20⬚C. (From Ref. 18.)



61

which corresponds to the surface-stable conformation (equilibrium adsorption conditions) is relevant. All the thermodynamic data discussed relate to the adsorption systems where the eluent composition is maintained the same during the chromatographic process (isocratic elution). For gradient elution, which is of special importance for modern HPLC, other thermodynamic treatment is necessary. One of the basic relationships used for the description in linear gradient elution can be expressed as follows: ¯ log k¯ = log k0 ⫺ S⌿

(6)

¯ is the value of the volume fraction of the organic solvent (or the median where ⌿ organic mole fraction) as the solute band passes the center of the column, k¯ is the ¯ = 0, and S is a pamedian capacity factor, k0 is the extrapolated value of k¯ for ⌿ rameter related to the magnitude of contact surface area between the solute and the stationary phase [36–40]. The median capacity factor k¯ thus substitutes k⬘ that is used in the case of isocratic elution. One example of the thermodynamic characterization of gradient elution systems was presented in a paper by Purcel et al. [37] where oligopeptides (up to nine amino acid residues) related to human growth hormone were chromatographed on octadecyl silica stationary phases. Negative ⌬H⬚ values (Table 2) derived from the Van’t Hoff plots indicate that heat is released upon the adsorption. At the same time, the ⌬S⬚ values found were in some cases negative (⫺32.8 to ⫺122.2 J mol⫺1 K⫺1 for the open-chain (␣- and ␤-linked peptides) and in some cases positive (2.6 J mol⫺1 K⫺1 for Phe5 and 3.8 J mol⫺1 K⫺1 for the imide, both of which exist as helices in solution), depending on flexibility of the solute molecules. The negative ⌬S⬚ values are indicative of an increased ordering of the overall system, and suggest that the open-chain peptides exist in more flexible conformation in solution than on stationary phase, whereas the helical structure is more rigid and constrained [37]. Note, that for the ␤-linked Leu-Ser-Arg-Leu-Phe-Glu-Asn-Ala-Gly, two sets of the thermodynamic parameters (Table 2) have been derived from the Van’t Hoff plot, log k versus 1/T, having two parts of different slopes which correspond to two different conformations at 5–65⬚C and 65–75⬚C. This peptide is composed of nine amino acid residues, and conformational transitions became even more expressed upon lengthening the chain. This is conventionally demonstrated by the plots of log

TABLE 2 Thermodynamic Data for Peptides Chromatographed on the C18 Column Using Gradient Elution Peptide

⌬H ⬚ (kJ mol⫺1)

⌬S ⬚ (J mol⫺1 K⫺1)

Leu-Ser-Arg-Leu-Phe-Asp-Asn-Ala (imide) Leu-Phe-Asp-Asn-Ala-Gly (␣) Leu-Ser-Arg-Leu-Phe-Glu-Asn-Ala-Gly (␤)

⫺3.3 ⫺8.7 ⫺35.4 ⫺6.1 ⫺3.3

3.8 ⫺39.7 ⫺122.2 ⫺32.8 2.6

Phe5 a

65–75⬚C. 5–65⬚C. Source: Ref. 37. b


⫾ ⫾ ⫾ ⫾ ⫾

0.6 0.3 8.9a 1.0b 0.6

⫾ ⫾ ⫾ ⫾ ⫾

0.6 1.2 30.6a 5.5b 1.1

62

Basiuk

k0 (or alternatively S, the contact area parameter) versus temperature (⬚C) which, however, cannot be used to derive ⌬H ⬚ and ⌬S⬚. The small peptide models Nacetylphenylalanine ethyl ester and N-acetyltryptophanamide exhibit a uniform decrease in log k0 with increase in temperature (Fig. 14a) and therefore no structural perturbations. Phe5, which exists as a helix in solution and has greater conformational flexibility, displays minor fluctuations with temperature (Fig. 14a). For bombesin, composed of 14 amino acid residues, a transition in the range of 5–25⬚C, corresponding to two different interactive structures, is observed on the C18 stationary phase. (On the C4 phase such transition proceeds at much higher temperatures, as shown in Fig. 14b.) A behavior essentially similar to that of bombesin was found for bovine insulin A chain (21 amino acid residues), whereas the B chain (30 amino acid residues) demonstrated multistep conformational changes within the whole temperature range of 5–85⬚C remnant of the behavior of the parent bovine insulin (Fig. 14c). Thus the thermodynamic characterization for the adsorption of flexible peptide molecules displaying conformational transitions should be very complicated since each conformation has its own ⌬H⬚ and ⌬S⬚ values. A detailed thermodynamic treatment of such a nonlinear Van’t Hoff behavior on hydrophobic silicas was exemplified for biologically active peptides bombesin, ␤-endorphin, and glucagon, as well as for some synthetic peptides, in Refs. 38, 41, and 42. C.

Cyclic Peptides

Scarce data are available so far on the adsorption thermodynamics of cyclic peptides. The only example studied using the HPLC technique was a series of cyclic dipeptides (or piperazine-2,5-diones) derived from ␣-alkyl amino acids adsorbed on bare silica from pure water [17,18]. Contrary to amino acid and linear dipeptide adsorption under the same conditions, most piperazinediones exhibit values of K > 1 and ⌬G⬚ < 0 (positive excess adsorption); the exceptions are cyclo-Gly2, cyclo-Gly-DL-Ala, and diastereomeric cyclo-Ala2 (K < 1 and ⌬G⬚ > 0), as well as cyclo-Aib2 (K = 1 and ⌬G⬚ = 0). Plotting ⌬G⬚ versus the number of aliphatic carbon atoms (nC), we obtained a linear trend (Fig. 15) similar to those shown for aliphatic bifunctional amino acids (Fig. 3) and related dipeptides (Fig 11). In the present case, the increment in ⌬G ⬚ for one aliphatic carbon atom is ca. ⫺510 J mol⫺1, i.e., adsorbability increases more significantly when increasing the size of the ␣-substituent than that for amino acids and linear dipeptides (see Sections III.A and B). Comparison of the ⌬G⬚ values found for piperazinediones and related linear compounds shows that the cyclization of the latter into the former results in ⌬G ⬚ decrease by ⫺70 J mol⫺1 for Gly-Gly → cyclo-Gly2; by ⫺180 J mol⫺1 for DL-Ala-DL-Ala → cyclo-(DL-Ala)2; and by ⫺990 J mol⫺1 for L-Val-L-Val → cyclo-(L-Val)2. In other words, amino and carboxylic groups reduce adsorbability of amino acids and related peptides on bare silica not only being present in the ␣-substituent: this conclusion can be applied to the terminal groups as well. D.

Nucleic Acid Bases

Despite the fact that many papers have been published on HPLC resolution of purine and pyrimidine nucleic acid bases, they have not focused on the thermodynamic aspects. Much retention data have been generated (e.g., Refs. 43 and 44) using © 2003 by Marcel Dekker, Inc.


63

FIG. 14 Dependence of log k0 on temperature for (a) penta-L-phenylalanine (●), N-acetylphenylalanine ethyl ester (䡩) and N-acetyltryptophanamide (䉮) chromatographed on C18 silica; (b) bombesin chromatographed on C18 (䡩) and C4 (●) silica; (c) bovine insulin (●) and its A(䊲) and B-chain (䡲) chromatographed on C18 silica. For the details of gradient elution, see


64

Basiuk

FIG. 15 Plot of ⌬G ⬚ versus the number of aliphatic carbon atoms (nC) in the piperazine2,5-dione molecules for the adsorption on bare silica from water. Temperature 19⬚C. (From Ref. 18.)

various silica stationary phases (bare silica, C2, C8, and C18, of different surface and porosity properties) and solvent systems; however, they were not treated to derive the thermodynamic characteristics. Some qualitative comments on the relative behavior of the solutes were provided. In particular, in the normal-phase chromatography (bare silica and mixed dichloromethane/methanol/aqueous eluents) [43], contrary to the reversed-phase mode, methyl substituents generally decrease the retention parameters; thus the adsorbability also decreases (⌬G ⬚ increases). For the reversedphase sorbents, the retention time is longer (and adsorption is correspondingly higher) on C18 than on C8; at the same time no significant differences have been found comparing the C2 and C8 materials [44]. Apparently, the brief communication by Kazakevich and El’tekov [45] remains after more than a decade the only work containing direct thermodynamic estimates for nucleic base adsorption. Using the chromatographic method, the ⌬H⬚ values have been determined for the adsorption of adenine, thymine, uracil, and cytosine on phenyl silica from solution of acetonitrile in water (20 vol%): 32.2, 15.9, 4.2, and 6.7 kJ mol⫺1, respectively. Driving forces for the purine adsorption on C18 and CN silicas from acetonitrile–water solutions were analyzed (in terms of chromatographic retention only, without thermodynamic estimates) by El’tekova and El’tekov [46], taking the parent purine, adenine, guanine, and xanthine as test adsorbates. Electrostatic, dispersion, dipole–dipole, and hydrophobic forces all contribute to the adsorption process. In particular, the dispersion interactions contribute more significantly on the hydrophobic C18 surface than on the CN silica, giving rise to 1.5 to 2.5-fold increase in the capacity factors. On the other hand, the presence of nitrile groups on the latter silica increases the contribution from dipole–dipole interactions. The purine structure has a pronounced effect on the retention characteristics as well. © 2003 by Marcel Dekker, Inc.


E.

65

Nucleosides

The situation outlined in the first paragraph of the previous section can be also applied to nucleosides. From the results of the HPLC retention studies one can conclude that ribonucleosides are systematically adsorbed stronger than their parent bases and the corresponding 2⬘-deoxyribonucleosides when chromatographed in the normal-phase mode [43], although no direct thermodynamic data have been presented. Some results have been afforded using the static (batch) approach [17,47]. Figure 16 shows that the isotherms for adenosine and inosine adsorption on bare silica from water have essentially the same shape (Langmuir-type), with about the same maximum adsorption capacity (ca. 45 nmol m⫺2) and ⌬G ⬚ values (⫺14.2 and ⫺13.5 kJ mol⫺1, respectively). Since the evident structural difference, i.e., the amino group in adenosine against the oxo group in inosine, does not influence noticeably the adsorption properties, one can assume that the purine nuclei do not contact the silica surface and, therefore, the carbohydrate moiety is responsible for the adsorption process. This assumption can be supported by analogous results obtained for the adenosine and inosine adsorption on ␥-aminopropyl silica from water [47], though the corresponding ⌬G⬚ values were found to be ⫺12.9 and ⫺11.5 kJ mol⫺1, respectively (Fig. 17), i.e., a little higher than in the case of bare silica. F.

Nucleotides

In the same works [17,47], the adsorption isotherms were measured for adenosine5⬘-triphosphate and uridine-5⬘-triphosphate on bare and ␥-aminopropyl silica from water. On bare silica, the nucleotides had lower values of maximal adsorption capacity (31 and 35 nmol m⫺2 for UTP and ATP, respectively) than the nucleosides (Fig. 16). This fact can be explained by the substantially bigger size of the nucleoside triposphate anions compared to that for the nucleosides by themselves, and their mutual electrostatic repulsion in the adsorbed phase. The calculated free energy changes are ⫺11.9 kJ mol⫺1 (ATP) and ⫺13.9 kJ mol⫺1 (UTP). For the case of ␥-aminopropyl silica (Fig. 17), the results are quite different. The maximum adsorption capacity for ATP, 140 nmol m⫺2, despite the difference in the molecular size, is almost four times higher than the corresponding value for the parent nucleoside, 36 nmol m⫺2. A big difference in their ⌬G ⬚ values was also found: ⫺17.6 versus ⫺12.9 kJ mol⫺1, respectively. Strong electrostatic attraction of the triphosphate anions to the surface aminopropyl groups is the most obvious reason for both phenomena. One could expect that UTP, due to a smaller molecular size, would have even higher adsorption capacity than ATP has. Actually the reverse effect is observed, which is hard to explain. As regards ⌬G ⬚, the value found for UTP is ⫺18.2 kJ mol⫺1, i.e., slightly lower than for ATP. One should note that studying these adsorption systems with nucleosides and nucleotides using the HPLC approach was inconvenient. The compounds were strongly retained on the bare silica column when eluted with pure water and gave very diffuse, poorly recognizable peaks [17,47]. Nevertheless, for buffer solutions with a higher ionic strength and organic additives typically used in HPLC (see, e.g., Ref. 44 and references therein), the dynamic method should be indispensable. Nucleotides, and especially the diphosphates and triphosphates, are rather ‘‘fragile’’ compounds and require certain precautions in their handling. It is best to © 2003 by Marcel Dekker, Inc.

66

Basiuk

FIG. 16 ⌬G ⬚ values determined from the isotherm initial parts for adsorption of adenosine, inosine, adenosine-5⬘-triphosphate, and uridine-5⬘-triphosphate on bare silica from water. Detection: UV at 270 nm. Temperature 15⬚C. (From Ref. 17.)

FIG. 17 ⌬G ⬚ values determined from the isotherm initial parts for adsorption of adenosine, inosine, adenosine-5⬘-triphosphate, and uridine-5⬘-triphosphate on ␥-aminopropyl silica from water. Detection: UV at 270 nm. Temperature 15⬚C. (From Ref. 47.)



67

buffer their solutions (unless unbuffered systems are of special interest) and to not expose the latter for a long time at even ambient temperatures. Otherwise one cannot be sure that after awhile the triphosphate remains triphosphate, and not monophosphate or diphosphate. What can result is clear: the adsorption isotherm and other characteristics will relate to the decomposition product(s). Take ATP as an example. The hydrolysis of its phosphate linkages can produce ADP or AMP and one or two phosphate anions. If UV absorption is used to determine equilibrium concentrations of the nucleotide in solution, an isotherm remnant of the ATP and adenosine isotherms should be obtained; adenine nuclei are detected, the total amount of which in the system remains the same. But if phosphate moieties are measured, the equilibrium concentration might appear to be overestimated by up to three times (at the extreme, one AMP molecule plus two phosphate anions). In particular, the method of phosphate ionometry was used in Ref. 48 to study the adsorption of biochemically significant phosphates on bare silica from unbuffered solutions. Apparently only the hydrolysis of phosphate linkages was the reason that ATP and ADP were found to be adsorbed very weakly, practically not adsorbed at all, strikingly differing from AMP (Fig. 18; for easy comparison, the scale has been adapted to that used in Figs. 16 and 17). No direct data are available on the thermodynamic behavior of oligonucleotides, beginning with dinucleotides. From the HPLC retention data and simple chromatograms presented elsewhere (e.g., Refs. 49 and 50), one can expect essentially similar regularities to those found for oligopeptides (Section III.B), at least to some limit of the chain length. Results of a recent study of the adsorption of fluorescently tagged oligonucleotide 5⬘-GTC AAG GCT GCC CAA TTT GAG-3⬘, which encodes a region of the human FcgRIIA gene, on ␥-aminopropyl–derivatized porous glass from phosphate buffer of pH 7.4 [51] suggest that this oligonucleotide (21 nucleotides long) still has a rigid rod structure, lying upon adsorption along its long axis parallel to the surface. The adsorption isotherm (Fig. 19) is a Langmuir isotherm, as in the case of the triphosphates (Figs. 16 and 17). At the same time, a free energy change calculated for the oligonucleotide (⫺10.4 kJ mol⫺1; Fig. 19) appears to be much higher than the ⌬G ⬚ values for ATP and UTP adsorption on the same ␥aminopropyl surface (Fig. 17) due to the strong effect of the phosphate buffer. In

FIG. 18 Isotherms of AMP, ADP, and ATP adsorption on bare silica from unbuffered aqueous solutions as measured by phosphate ionometry. (From Ref. 48.)


68

Basiuk

FIG. 19 Adsorption isotherm of fluorescently tagged oligonucleotide 5⬘-GTC AAG GCT GCC CAA TTT GAG-3⬘ on ␥-aminopropyl–derivatized porous glass from phosphate buffer of pH 7.4. (From Ref. 51.)

this particular case the adsorption was reversible due to high ionic strength. On the other hand, one can expect irreversible adsorption of the same oligonucleotide from solutions of low ionic strength (e.g., water as the simplest case) due to the formation of polyphosphate salts with the surface ␥-aminopropyl groups. Adsorption reversibility should necessarily be verified prior to determining equilibrium constants from the initial isotherm slopes and then deriving any thermodynamic data, since in the case of irreversible binding the conventional Langmuir model cannot be applied (e.g., for long-chain DNA adsorption [52]).

IV.

CONCLUSIONS AND FUTURE PROSPECTS

It is clearly seen that the main body of thermodynamic data for biomolecular adsorption on silica has been afforded for amino acid adsorption on the surface of chromatographic sorbents. This is quite understandable taking into account the role of HPLC in the modern practice of amino acid analysis and separation. Much thermodynamic data have been derived (basically for the adsorption on bare and octadecyl silica from water). As an example of structure–property relationships, dependence of the Gibbs free energy on the amino acid structure (carbon number) has been shown. Also, compensation behavior has been shown to be applicable for amino acid series adsorption on bare and octadecyl silica. As a whole, the HPLC approach has proved to be the most useful to study thermodynamics of amino acid adsorption on silicas; some data, nevertheless, can be afforded from adsorption isotherms obtained from batch measurements. At the same time, some descrepancies can be revealed in the thermodynamic data presented by different groups, thus indicating the necessity of their revalidation. Thermodynamic characterization of small peptide adsorption is so far at a less advanced level. Perhaps the most important generalization remains a linear dependence of the retention values, equilibrium constants and free energy changes on the © 2003 by Marcel Dekker, Inc.


69

number of amino acid residues in homopeptides, which is maintained to a certain limit of peptide chain length. Naturally, there is a great similarity in adsorption behavior of amino acids and related peptides, which can be found in the dependence of ⌬G⬚ on the carbon number of peptide molecules. Regarding nucleic acid components, only scarce data have been reported, derived mostly from the batch-measured adsorption isotherms. There is much to do in this area; of special interest would be systematic studies on how adsorption characteristics change from pyrimidine and purine bases to related nucleosides, nucleotides, and further to oligonucleotides. To conclude, studying the relative behavior within a class of compounds can be much more useful to understand the interfacial behavior of biomonomers and biopolymers than attempts to estimate particular thermodynamic values for a given compound, which are often insufficiently precise and meaningful only for a given adsorption system.

REFERENCES 1. 2.

3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

A. Fallon, R. F. G. Booth, L. D. Bell, Applications of HPLC in Biochemistry. Elsevier, Amsterdam, 1987. C. T. Mant and R. S. Hodges, eds., High Performance Liquid Chromatography of Peptides and Proteins: Separation, Analysis and Conformation. CRC Press, Boca Raton, FL, 1991. E. D. Katz, ed., High Performance Liquid Chromatography: Principles and Methods in Biotechnology. Wiley, Chichester, 1996. G. D. Parfitt, C. H. Rochester, eds., Adsorption from Solution at the Solid/Liquid Interface. Academic Press, London, 1983. H.-H. Kohler. In: B. Dobias, ed., Coagulation and Flocculation. Marcel Dekker, New York, 1993, pp. 1–36. J. Torrent, F. Sanz. J. Electroanal. Chem. 286:207–215 (1990). C. H. Giles, T. H. MacEwan, S. N. Nakhwa, D. Smith. J. Chem. Soc. 3973–3993 (1960). C. H. Giles, D. Smith, A. Huitson. J. Colloid Interface Sci. 47:755–765 (1974). R. P. W. Scott. In: P. R. Brown, R. A. Hartwick, eds., High Performance Liquid Chromatography. Wiley, New York, 1989, pp. 117–143. B. Karger. In: J. J. Kirkland, ed., Modern Practice of Liquid Chromatography. Wiley, New York, 1971, Chap 1. K. S. Yun, C. Zhu, J. F. Parcher. Anal. Chem. 67:613–619 (1995). N. A. Eltekova, D. Berek, I. Novak. Zh. Fiz. Khim. (Russian) 63:2675–2678 (1989). Y. V. Kazakevich, H. M. NcNair. J. Chromatogr. Sci. 31:317–322 (1993). S. Ahuja, ed., Chiral Separations by Liquid Chromatography. American Chemical Society, Washington, D.C, 1991. I. Molna´r, C. Horva´th. J. Chromatogr. 142:623–640 (1977). V. A. Basiuk, T. Yu. Gromovoy, E. G. Khil’chevskaya. Polish J. Chem. 68:777–781 (1994). V. A. Basiuk, T. Yu. Gromovoy, E. G. Khil’chevskaya. Origins Life Evol. Biosphere 25: 375–393 (1995). V. A. Basiuk, T. Yu. Gromovoy. Adsorption 2:145–152 (1996). M. S. Uddin, K. Hidajat, C.-B. Ching. Ind. Eng. Chem. Res. 29:647–651 (1990). V. A. Basiuk, T. Yu. Gromovoy. Colloids Surfaces A 118:127–140 (1996). W. Melander, D. E. Campbell, C. Horva´th. J. Chromatogr. 158:215–225 (1978).


70

Basiuk

22. 23. 24. 25.

A. Vailaya, C. Horva´th, J. Phys. Chem. 100:2447–2455 (1996). R. R. Krug, W. G. Hunter, R. A. Grieger. J. Phys. Chem. 80:2335–2341 (1976). R. R. Krug, W. G. Hunter, R. G. Grieger. J. Phys. Chem. 80:2341–2351 (1976). J. P. Greenstein, M. Winitz. Chemistry of the Amino Acids, Vol 1. RE Kreiger Publishing, Malabar, 1986, p. 486. M. C. J. Wilce, M.-I. Aguilar, M. T. W. Hearn. Anal. Chem. 67:1210–1219 (1995). A. H. Rodgers, M. G. Khaledi. Anal. Chem. 66:327–334 (1994). V. A. Davankov, A. S. Bochkov, Yu. P. Belov. J. Chromatogr. 218:547–557 (1981). V. A. Basiuk, A. A. Chuiko. J. Chromatogr. 521:29–42 (1990). T. C. Pochapsky, Q. Gopen. Protein Sci. 1:786–795 (1992). L. T. Kubota, A. Gambero, A. S. Santos, J. M. Granjeiro. J. Colloid. Interface Sci. 183: 453–457 (1996). J. L. Meek, Z. L. Rossetti. J. Chromatogr. 211:15–28 (1981). M. T. W. Hearn, B. Grego. J. Chromatogr. 296:309–319 (1984). C. T. Mant, T. W. L. Burke, J. A. Black, R. S. Hodges. J. Chromatogr. 458:193–205 (1988). K. Benedek. J. Chromatogr. 646:91–98 (1993). A. W. Purcell, M. I. Aguilar, M. T. W. Hearn. J. Chromatogr. A 711:71–79 (1995). A. W. Purcell, M. I. Aguilar, M. T. W. Hearn. J. Chromatogr. 476:125–133 (1989). A. W. Purcell, M. I. Aguilar, M. T. W. Hearn. J. Chromatogr. 593:103–117 (1992). A. W. Purcell, M.-I. Aguilar, M. T. W. Hearn. Anal. Chem. 65:3038–3047 (1993). B. de Collongue-Poyet, C. Vidal-Madjar, B. Sebille, K. K. Unger. J. Chromatogr. B 664: 155–161 (1995). A. W. Purcell, G. L. Zhao, M. I. Aguilar, M. T. W. Hearn. J. Chromatogr. 852:43–57 (1999). R. I. Boysen, Y. Wang, H. H. Keah, M. T. W. Hearn. Biophys. Chem. 77:79–97 (1999). M. Ryba, J. Berańek. J. Chromatogr. 211:337–346 (1981). A. Rizzi, H. R. M. Lang. J. Chromatogr. 331:33–45 (1985). Yu. V. Kazakevich, Yu. A. El’tekov. Russ. J. Phys. Chem. 61:233–234 (1987). N. A. El’tekova, Yu. A. El’tekov. Russ. J. Phys. Chem. 74:1324–1329 (2000). V. A. Basiuk, T. Yu. Gromovoy, E. G. Khil’chevskaya. Polish J. Chem. 69:127–131 (1995). K. Hamdani, K. L. Cheng. Colloids Surfaces 63:29–31 (1992). M. Colpan, D. Riesner. J. Chromatogr. 296:339–353 (1984). R. Bischoff, L. W. McLaughlin. J. Chromatogr. 296:329–337 (1984). V. Chan, D. J. Graves, P. Fortina, S. E. McKenzie. Langmuir 13:320–329 (1997). K. A. Melzak, C. S. Sherwood, R. F. B. Turner, C. A. Haynes. J. Colloid Interface Sci. 181:635–644 (1996).

26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52.


4 Quantitative Modeling of Protein Adsorption CHARLES M. ROTH Massachusetts General Hospital, Harvard Medical School, and Shriners Burns Hospital, Boston, Massachusetts, U.S.A. ABRAHAM M. LENHOFF Newark, Delaware, U.S.A.

I.

University of Delaware,

INTRODUCTION

The general principles that have been deduced from experimental observations of protein adsorption are covered in several major review articles [1–3]. Our focus here is more specific, namely, on the development of a quantitative understanding of the mechanistic bases driving proteins to accumulate at solid–liquid interfaces and on the ability to predict extent of adsorption from physicochemical properties of the protein, aqueous solvent, and sorbent. Such a capability would allow rational approaches to be developed toward the solution of problems involving protein adsorption. For example, a detailed understanding of how the properties of protein and surface are manifested in their adsorption behavior would aid in the design or selection of materials that selectively adsorb proteins onto micropatterned surfaces for tissue engineering and biosensor applications. For protein chromatography, a quantitative understanding of the extent of adsorption as a function of protein and adsorbent properties and solvent strength would improve the process of selecting adsorbents and conditions most amenable to particular protein separations. One key factor that distinguishes protein adsorption to biomaterials from adsorption to chromatographic stationary phases is the extent to which the adsorption is reversible. The adsorption of blood proteins, which often possess rather flexible structures, to the polymeric substrates usually used as biomaterials has been found to be irreversible in many cases; often conformational change occurring over long contact times is associated with it [1–3]. On the other hand, the ability to elute proteins from a chromatographic column, particularly in isocratic operation, demonstrates that protein adsorption in these systems is essentially reversible. Because the current level of quantitative understanding regarding protein adsorption is not very advanced, we focus here on systems of the latter type. The understanding of protein adsorption that does exist currently is in the form of basic thermodynamic concepts, such as a change in the Gibbs free energy upon adsorption [4]; dealing with irreversible adsorption within a thermodynamic framework is beyond the scope © 2003 by Marcel Dekker, Inc.

72

Roth and Lenhoff

of present capabilities. In practical terms, this excludes adsorption in which conformational change and/or strong hydrophobic interactions are significant. Nonetheless, understanding of the mechanisms driving protein adsorption will be necessary in understanding these more complex situations as well. Here, we focus on aqueous systems for which protein adsorption occurs as the net result of protein molecular interactions with the adsorbent relative to those of the protein and sorbent with their solvent. Electrostatic interactions are very important in this case, as many of the residues of a protein are ionized at any given pH. Dispersion, or van der Waals, forces are almost always attractive and so form a weak but nonnegligible contribution to adsorption. Solvation and steric forces may be significant as well. These different contributions are discussed in more detail in Section II. Given the contribution of each of these mechanisms to the interactions that govern adsorption, one needs a framework to combine the interaction energies or forces that are computed with information regarding the formation and structure of the adsorbed layer. Two classes of approach exist: one in which a thermodynamic framework utilizes configurational averaging to estimate bulk properties and a second in which the adsorption process is simulated from the interaction potentials (forces and energies). In either case it is possible to obtain descriptions of adsorption that can be related to protein and adsorbent properties and can thus be applied with some predictive capability. The descriptions sought are limited to ones of the adsorbed amount, which is usually of principal experimental interest. In Section III, these mechanistic models for predicting adsorption behavior are reviewed. Early models for protein adsorption did not account for structural aspects of the protein or sorbent, but instead treated the protein as a monofunctional chemical entity and the sorbent as a collection of chemical sites. In this approach, isotherms are generated via mass action equations from the proposed reaction, and slight variations result from altering the description of the reaction stoichiometry. However, for mechanistic understanding and a priori prediction of adsorption behavior, more detailed models accounting for structural aspects of the protein and/or surface are required. Protein structures and their associated behavior are quite complex. The amino acids that comprise proteins span a wide range of chemical functions, and a particular protein is likely to contain each of the 20 major amino acids somewhere within it. Therefore, a folded protein is likely to have different regions that are hydrophilic or hydrophobic; positively charged, negatively charged, or neutral; internal or exposed to solvent; and functional or structural. Because of this heterogeneity at the amino acid residue level, chemical treatments of protein adsorption are inadequate. With recent advances in computational capabilities, incorporation of atomic information for both proteins and surfaces into mechanistic models of protein adsorption has begun to be tractable in some cases. Molecular mechanics computations and molecular dynamics simulations have become commonplace in modeling smaller molecules in aqueous solvents, but their application to macromolecules, including proteins, is still limited by computational resources. Nonetheless, atomistic models of proteins and polymer surfaces have been used in a few cases to calculate adsorption energies [5–7] and to investigate dynamic [8] and conformational [9] effects involving the adsorption of proteins onto polymer surfaces. © 2003 by Marcel Dekker, Inc.

Quantitative Modeling of Protein Adsorption

73

These models have provided some insight into the relative contributions of various types of interactions (e.g., electrostatic, dispersion, solvation) to protein– polymer interactions, but the problem of simulating entire protein molecules and their adsorbents with sufficient water and ions to be realistic remains too demanding for extensive study. Consequently, a number of approaches have been applied in which the protein molecule is treated atomistically, but continuum approximations are made for the solvent and/or adsorbent. Noinville et al. [10] used atomistic descriptions of protein and poly(vinylimidazole) polymer, the force field parameters of the latter determined via quantum mechanical computations. Solvent was included by modifying the intermolecular potentials (AMBER force field [11]) by a distancedependent dielectric constant and by an exponential screening term that accounts for the double-layer screening with increasing ionic strength. Others have also described protein and polymer atomistically, using interatomic potentials designed to account for the presence of intervening water [5]. The solvation energy absent from such a description was added by a fragment method based on aqueous–organic transfer free energies of constituent amino acids [6,7]. This approach has been further extended to account for the presence of grafted polymer chains on the adsorbent substrate [12]. Continuum, or colloidal, methods represent an alternative to atomistic approaches that are particularly advantageous in their ability to deal with the problems of solvent and electrolyte. The particle size over which a colloidal description is appropriate ranges from about a nanometer to a micrometer. Most proteins fall toward the lower end of this range but are certainly within it. Since the sizes are small, protein interaction energies are relatively small as well, and consequently proteins can be sensitive to minor changes in environment. Choosing a colloidal approach engenders a few assumptions. First, treating a protein molecule as a rigid particle makes it virtually impossible to account for conformational change in the adsorption process. While protein lability can be an important driving force for adsorption [13], predicting conformational changes is beyond the range of current capabilities. Second, colloidal approaches deal primarily with nonspecific interactions. Specificity is included only with respect to strengthening of dispersion interactions via complementarity of shape. For example, it is not possible to assess, from the point of view of particle–surface interactions, changes in the hydrogen-bonding state of protein, surface, or solvent. Third, all materials involved in the adsorption process are treated as continua. Consequently, properties of multifunctional adsorbents are taken into account only in an average or statistical sense. Other solutes are generally treated as part of the solvent medium, particularly ions, which are treated as point charges and lumped into the Debye screening parameter, which depends on the ion valence and concentration but not on its specific chemical type. Overall, the implication is that a colloidal description is good for describing the effect of nonspecific interactions on bulk association but is not useful for specific effects, strongly hydrophobic sorbents, or adsorption driven by conformational change. In the remainder of this chapter, we emphasize a molecular understanding of the governing interactions as a fundamental basis for analyzing and manipulating the bulk adsorption and chromatographic behavior of proteins. These interactions can be described from either atomistic or colloidal perspectives, and the choice of formulation depends on the extent of structural information and computational resources available. We review quantitative approaches to describing the molecular interactions © 2003 by Marcel Dekker, Inc.

74

Roth and Lenhoff

involved in protein adsorption in both atomistic and colloidal frameworks. Thereafter, we proceed to use these methods as the basis for calculating adsorption equilibrium, and isotherms. A suitable thermodynamic framework allows a quantitative model for interactions at the molecular (or, from the colloidal view, particulate) level to be used to make predictions or interpretations regarding bulk behavior, which can be tested experimentally. II.

PROTEIN–SURFACE MOLECULAR INTERACTIONS

A.

Electrostatics

At the atomic level, electrostatics are treated pairwise, with the potential energy Uij of each charge pair following from Coulomb’s law. Thus, the electrical energy Uelec of a configuration of atoms is Uelec =

冘 i<j

qi qj ␧r

(1)

where qi and qj are the charges on any two atoms i and j, ␧ is the dielectric permittivity of the intervening medium, and r is the distance between charge centers. The source of charges within the protein usually consists not just of the fixed charges resulting from titratable amino acid residues, but also of partial charges of the polypeptide backbone and side chains. The electrical energy contains contributions from both intramolecular and intermolecular Coulombic interactions; under the assumption of a rigid body, the intramolecular interactions can be omitted from the summation. Intermolecular interactions in aqueous media are strongly mediated by the presence of water and ions. In a dynamic simulation, these can be included explicitly; however, in a static calculation, approximations for their effects on interatomic charge interactions are introduced. The presence of water is treated by inclusion of the solvent dielectric constant between each pair of atoms in Eq. (1); sometimes a distancedependent value is used. The presence of electrolyte can be approximated by scaling the interactions in Eq. (1) by an exponentially decaying function of distance, with the Debye length as the decay length. The problems of intervening solvent and electrolyte are often not suitably addressed with the aforementioned approximations, and as a result continuum treatments for electrostatics have been developed. These generally build on the cavity dielectric model of Kirkwood [14], in which the protein is treated as a low-dielectric body immersed in a continuous electrolyte solution. The interior of the protein is governed by the Poisson equation: ⵜ2␾ = ⫺

␳ ␧

(2)

which relates the electrical potential ␾ in a region of dielectric permittivity ␧ to the distribution of charges ␳ in that region. For interactions in an electrolyte solution, the free ions are assumed to follow a Boltzmann distribution, leading to the Poisson– Boltzmann equation (PBE): ⵜ2␾ = ⫺

1 ␧

冘

冉冊

␳ oi zi e exp ⫺

i

zi e␾ kT

(3)

where k is the Boltzmann constant, T the temperature, zi the valency of ion species © 2003 by Marcel Dekker, Inc.


75

i, e the magnitude of the electronic charge, and ␳ oi the bulk concentration of species i. The linearized form of Eq. (3) (LPBE) is used frequently, and while linearization is strictly valid only for very low potentials, surprisingly good agreement between energies computed using the PBE and the LPBE has been found for a variety of double-layer calculations [15–18]. Solution of these coupled equations provides the electrostatic potential distribution throughout space, from which the electrostatic interaction energy is calculated by integration of the potential changes at each charge site [19,20] using an expression of the form 1 F= 2

冕

⭸P

1 ␴␾ dA ⫹ 2 e

冘 n

qk␾i(x k)

(4)

k =1

Here, the superscripts e and i denote exterior and interior potentials, ␴ is the charge density on a uniformly charged planar adsorbent surface, x k are the locations of the n charges, and ⭸P represents the planar surface. For a geometrically complex particle such as a protein, the governing electrostatic equations must be solved numerically. A number of programs for solution of these equations in and around protein molecules have been developed. Early implementations relied almost exclusively on the finite difference technique [21,22], but alternatives such as finite elements [23] and the boundary element method [18,20, 24–27] have been developed. While these programs are generally rather computer intensive, one calculation is able to give a wealth of information in terms of the electrostatic potential distribution everywhere in space. The growing literature on this subject has addressed many of the key issues in the model, including the boundary condition at the protein–solvent interface, intramolecular interactions among charged groups (pKa shifts) [28], and the dielectric constant of the protein [20,30]. For the continuum methods described, the electrostatic potential distribution and the free energy of interaction are both useful in explaining adsorption behavior. The electrostatic potential distribution alone has provided some qualitative explanations of protein adsorption orientation and layer structure. The potential distribution around ribonuclease A, along with its interaction energy, has been found to favor its adsorption to a negatively charged surface in a side-on orientation with its active site facing toward the surface [31]. Furthermore, since the opposite side of the ribonuclease molecule features a negative potential, an adjustment to an end-on orientation for close-packed adsorbed molecules could be stabilized by favorable electrostatic interactions. This proposed mechanism was used to explain observed structural changes in adsorbed layers of ribonuclease A [31] based on surface forces measurements and enzymatic activity changes. Explicit inclusion of a charged surface in electrostatic calculations [20] corroborated these conditions. In a similar manner, Haggerty and Lenhoff [32] used the electrostatic potential distribution around lysozyme in solution to propose a mechanism for electrostatically favored ordering of protein molecules adsorbed at a surface. While the potential distribution does not allow direct prediction of adsorption extent or adsorbed layer structure, these studies highlight the important role of electrostatics in protein adsorption and complement the quantitative models discussed later. Molecular detail in finding the free energy of interaction is also informative. One observation is that a striking dependence on orientation is possible, e.g., that some orientations result in favorable energies of interaction while other orientations exhibit unfavorable energies of interaction for proteins and surfaces of opposite net © 2003 by Marcel Dekker, Inc.

76

Roth and Lenhoff

charge. This can be seen in Fig. 1, where the electrostatic free energies as a function of orientation are presented for chymotrypsinogen A (net charge ⫹4 at pH 7) interacting with a uniformly charged anionic surface. A range of interaction energies exists, but in most orientations, chymotrypsinogen A is attracted to the negatively charged surface; such behavior is consistent with its positive net charge. Analogous behavior has been reported for the interaction of rat cytochrome b5 at pH 8, where it is negatively charged, with an array of model anion exchange ligands represented atomistically [33]. However, the calculated attractive energies in this system are enormous (up to ⫺58 kT), presumably due to the large protein net charge (⫺9.4 e) and surface charge density (10 ␮C/cm2) employed. The orientational dependence of the free energy of interaction has also been investigated in a system where the protein and sorbent are both positively charged. Ribonuclease A and cytochrome C have been observed to adsorb to anion exchange (positively charged) surfaces under conditions at which both proteins are positively charged [34]; this behavior has been ascribed to patches of negative electrostatic potential around the generally positively charged proteins. Asthagiri and Lenhoff [35] have computed interaction energies for these proteins in orientations most likely to produce this patch-controlled behavior. When the sorbent was modeled as a surface with uniform charge density, electrostatics computations were unable to reproduce the favorable interaction energies necessary for patch-controlled adsorption. However, when the sorbent charges were modeled as an array of discrete charges or as isolated low-dielectric charged spheres (approximating the topography of polymer chain end groups), favorable interaction energies did result from the model. Such a pronounced orientational dependence of the interaction energy is not a universal phenomenon. For lysozyme, the dependence of electrostatic energy on orientation (Fig. 2) is different from that for ribonuclease or chymotrypsinogen. Under the same conditions as presented for chymotrypsinogen A, lysozyme exhibits some dependence on orientation, with values ranging from ⫺1.6 to ⫺4.0 kT, but in all cases the electrostatic interaction is attractive. An explanation is provided by the fact

FIG. 1 Chymotrypsinogen A–surface electrostatic interaction energies as a function of ori˚ , ionic strength = 0.1 M, ␴ = ⫺4.6 ␮C/cm2. entation (␾, ␪ ). Conditions: gap = 7.66 A



77

FIG. 2 Lysozyme–surface electrostatic interaction energies as a function of orientation (␾, ˚ , pH 7, ionic strength = 0.1 M, ␴ = ⫺4.6 ␮C/cm2. ␪ ). Conditions: gap = 7.66 A

that the molecular charges of lysozyme are unusually evenly distributed [36]. One measure of the charge distribution is the dipole moment, which for lysozyme is significantly less than that for chymotrypsinogen A (approximately 70 versus 520 Debye at pH 7). The electrostatic interaction energy provides a means to assess the importance, relative to other intermolecular forces, of electrostatic interactions in protein adsorption, and often approximate determinations are adequate because of the high computational demand of protein molecular electrostatics computations. The sphere– plate geometry is most convenient, albeit a numerical solution is nontheless required. Exact enumeration of electrostatic potentials and energies for a sphere and plate are useful in capturing the essential physics of colloidal particle–surface electrostatic interactions. The main drawback in application to proteins is the neglect of the heterogeneity of the charge distribution within a protein molecule that is likely to favor certain configurations in protein–surface interactions. An intermediate representation of protein electrostatics can be made by incorporating the dipole moment of the charge distribution into the sphere model [37]. For proteins with a relatively symmetric charge distribution, such as lysozyme, the addition of the dipole moment is a relatively small perturbation and the interaction energies have a fairly narrow distribution, with the most favorable orientation being that in which the positive end of the dipole faces the surface (Fig. 3). This is in fairly good agreement with the protein computations of Fig. 2, but the sphere–dipole model underestimates the maximum free energy somewhat. For proteins with an asymmetric charge distribution, such as chymotrypsinogen A, the addition of the dipole moment to the charge distribution is a large perturbation, with the result that the distribution of energies is quite large and actually overestimates the range observed for the protein as a whole (Fig. 4). Nonetheless, incorporation of the dipole moment is relatively simple and seems to improve the accuracy of the characterization of electrostatic energies. Higher-order moments of the charge distribution could be incorporated as well. An alternative approach has been developed by Grant and Saville [38], who represented lysozyme as a sphere with the charge distribution © 2003 by Marcel Dekker, Inc.

78

Roth and Lenhoff

FIG. 3 Interaction energies for lysozyme, represented as a sphere with charges mimicking the protein’s monopole and dipole moments, as a function of dipole orientation, indicated by ˚ , pH 7 the angle ␪ between the dipole axis and the surface normal. Conditions: R = 15.6 A (net charge = ⫹8, dipole moment = 72 D), ionic strength = 0.1 M, ␴ = ⫺4.6 ␮C/cm2. (——) ˚ ; (⭈ ⭈ ⭈ ⭈) gap = 7.66 A ˚ ; (–⭈–⭈–⭈) gap = 2.15 A ˚. gap = 21.98 A

smeared onto a finite set of patches on its surface. This model also is able to produce orientation-dependent energies that match those calculated from detailed protein electrostatics semiquantitatively. For some applications, an approximate analytical solution to the sphere-plate electrostatics equations is used. The most commonly employed approximation to the double-layer interaction energy in colloid science is that provided by the linear superposition approximation (LSA). The LSA represents the potential at any point between two charged bodies as the sum of the unperturbed potentials of the interacting bodies. This approximation is essentially correct when the bodies are far apart, but is less valid as the double layers overlap more extensively. Furthermore, the scalings of energy with respect to distance, surface charge density, and to some extent ionic strength predicted by the LSA hold for the models employing structural details, given a particular protein-surface orientation [39]. An analytical LSA result for a sphere with a nonuniform charge distribution interacting with a charged plate has also been developed [97].

B.

van der Waals Interactions

van der Waals forces arise from fluctuations in atomic dipoles and the polarizability of constituent atoms within the given materials that are interacting. For an atomistic description of macromolecules, the interaction is described by the Lennard–Jones potential © 2003 by Marcel Dekker, Inc.


79

FIG. 4 Interaction energies for chymotrypsinogen A, represented as a sphere with charges mimicking the protein’s monopole and dipole moments, as a function of dipole orientation, indicated by the angle ␪ between the dipole axis and the surface normal. Conditions: R = ˚ , pH 7 (net charge = ⫹4, dipole moment = 516 D), ionic strength = 0.1 M, ␴ = ⫺4.6 19.3 A ˚ ; (⭈ ⭈ ⭈ ⭈) gap = 7.66 A ˚ ; (–⭈–⭈–⭈) gap = 2.15 A ˚. ␮C/cm2. (——) gap = 21.98 A

Uij =

冘冉 i<j

冊

Bij Cij ⫺ 6 r 12 r ij ij

(5)

where Cij is a constant dependent mainly on the polarizabilities of the two atoms or groups (London dispersion force), but also on their dipole moments (Keesom and Debye forces). The first term in Eq. (5), parameterized by its constant Bij , is the Born repulsion term, discussed in more detail in Section II.D. The total van der Waals energy of a discrete collection of atoms is formed by summation of the second term in Eq. (5) over all pairs of atoms ij, although multibody effects, such as those captured by the Lifshitz theory discussed subsequently, are not properly accounted for. Atomistic descriptions are suitable for very short range interactions, but are very demanding computationally for the same reason as discussed previously for electrostatic effects, namely, the need to include explicit water molecules. Thus, at somewhat longer range a continuum colloidal approach is usually adopted. The treatment of van der Waals interactions in colloids is usually based on the approach of Hamaker [40], who described the total interaction of two bodies by integrating the characteristic 1/r 6 dependence of gaseous dispersion interactions over the volume of the interacting bodies. A major consequence of this approach is that the van der Waals interaction energy depends on the product of a material property, named the Hamaker constant A, and a function of geometry. This separation is a direct consequence of the assumption of pairwise additivity inherent in the Hamaker approach. The Lifshitz theory [41] is a rigorous alternative, based on quantum electrodynamics, © 2003 by Marcel Dekker, Inc.

80

Roth and Lenhoff

that accounts for the many-body interactions within interacting continuous bodies. Calculations for the only two tractable geometries—sphere/sphere and plane/plane —suggest that the Hamaker approach accounts correctly for the geometric dependence at close separations [42,43]. The Hamaker integration has been performed for a variety of geometries [44]. The result most often applied to protein adsorption models is that for a sphere and a plate, ⌬FvdW = ⫺

A132 6

冋

R R ⫹ ⫹ ln z 2R ⫹ z

冉

冊册

z 2R ⫹ z

(6)

where R is the sphere radius and z the separation distance between the sphere and the plate. It can be readily seen that the van der Waals interaction energy increases with the size of the spherical particle and decays with increasing separation distance. Mathematically, the interaction becomes infinitely attractive as the sphere and plate touch (z = 0), but short-range repulsive forces prevent this from occurring in practice, and the validity of the continuum representation is questionable here in any event. The van der Waals interaction is effective only at close range, but in that range it can be quite significant. The consequence for protein adsorption is that van der Waals interactions are probably less important in drawing a protein to a surface than in binding it once it arrives at the surface. Because of the complex structure of a protein molecule, most analyses of van der Waals interactions involving proteins assume the simplified geometry of a sphere [37,45] or even a plate [46,47]. The effect of protein geometry on van der Waals interactions has been investigated [48]. It was found that the interaction of proteins with planar surfaces is generally disfavored relative to that of spheres and planes because the roughness of the protein surface prevents most of its volume from being close enough to the surface to interact with it [48]. Certain orientations, however, are able to bring a fraction of the protein volume comparable to that of an equivalent sphere near to the surface. The converse, however, is also true: when two bodies have complementary shapes, their van der Waals attraction is enhanced [48]. For application to protein adsorption, the sorbent is frequently not well characterized in terms of molecular topography, and the exact value of the van der Waals contribution is difficult to ascertain. In order to apply the approach of Hamaker quantitatively, it is necessary to estimate the material property describing the mutual polarizability of protein and surface—the Hamaker constant. The rigorous framework of Lifshitz theory [4] allows the Hamaker constant to be estimated from the frequency-dependent dielectric spectra of the materials involved (including solvent). In particular, the adsorption spectra, along with estimates of the static dielectric constant and the refractive index in the visible region for each of the interacting materials as well as the intervening solvent, are sufficient to obtain a fairly accurate value of the Hamaker constant [49,50]. Relatively little work has been devoted to obtaining the dielectric data required for accurate determination of Hamaker constants, especially for studying proteins. The only complete set of dielectric data is that of Inagaki et al. [51], who measured the dielectric response of bovine serum albumin (BSA) over a wide frequency range. Since most proteins are composed of the same 20 amino acids, it can reasonably be assumed that the BSA data are representative of proteins in general [48]. Using these © 2003 by Marcel Dekker, Inc.


81

data, the Hamaker constant for two proteins interacting through water is found to be about 3.1 kT [48]. In aqueous electrolyte solution, the zero-frequency contribution is partially screened. The result is a reduction in the Hamaker constant for two bodies with an intervening aqueous electrolyte of about 0.75 kT; this reduction is independent of the nature of the materials, as long as their static dielectric constants are significantly less than that of water. For the analysis of protein adsorption, Hamaker constants for the interaction of BSA with a selection of materials have been evaluated [48]. The Hamaker constant for protein interactions with other proteins is higher than that for protein interactions with most of the materials on which they are likely to adsorb [48,52]. C.

Solvation Forces

Solvation, or hydration, forces are thought to arise from perturbations in the molecular structure of water in the vicinity of dissolved solutes or surfaces that can be either hydrophobic or hydrophilic in nature. We concentrate here on hydrophobic phenomena, which are of major importance for many protein adsorption situations. By coming together, two hydrophobic forces can reduce the total surface area exposed to solvent. The details of solvation interactions are still incompletely understood, and the term ‘‘hydrophobic effect’’ is probably an overused catch phrase. Nonetheless, there is experimental evidence implicating solvent-induced forces [53,54]. Since the amount of protein adsorbed generally increases with surface hydrophobicity [2], with hydrophobicity usually measured by water contact angle, it is sometimes stated that hydrophobic interactions provide the primary driving force for protein adsorption [55]. While this may well be the case for many hydrophobic surfaces such as those used in biomaterials studies, it is probably an overgeneralization. Nonetheless, the hydrophobic driving force is certainly an important one, and the poor understanding of its nature is a major factor in much of the uncertainty and controversy in protein adsorption studies. In view of the molecular origins of the hydrophobic effect, much effort has been expended on calculations and simulations involving water structure at the molecular level. In molecular dynamics, no hydrophobic force is necessary to produce solvation effects; they are a natural consequence of water as a solvent. Descriptions of solvation effects at the colloidal level are more difficult to model accurately, but are straightforward to implement into quantitative models of protein adsorption. There exist a number of macroscopic theories regarding solvation effects [56,57]. These have in large part been linked to what is commonly called hydrophobic hydration, which is generally characterized for a given species by the free energy of transferring that species from a nonpolar solvent to water. For hydrocarbons, the free energy is positive, reflecting the preference for the nonpolar solvent. The transfer free energy is correlated with the surface area of the solute, with approximately 85– ˚ 2 being contributed for nonpolar amino acid side chains [58]. More 180 J/mol A detailed breakdowns into atomic [59] or larger fragments [60] have also been suggested. Indeed, the latter approach has been used to estimate the hydrophobic contribution to the adsorption energy for proteins at polymer surfaces [6]. The hydrophobic interaction energy ⌬Fhphob in this approach is computed by a summation of terms: © 2003 by Marcel Dekker, Inc.

82

Roth and Lenhoff

⌬Fhphob =

冘

RT ln Pk

(7)

k

where Pk is a partition coefficient for the atom or fragment. The validity of applying partition coefficients measured for amino acids to fragments of a protein remains unclear, however. Furthermore, the dependence on separation distance must be added by an empirical exponential decay function [7]. Hydrophobic interactions have been measured in surface forces apparatus experiments [61,62], but the applicability of the results to protein interactions is questionable for at least two reasons. First, the force measurements are based on extended surfaces, and the variation with curvature remains to be clarified. Second, the heterogeneity of the protein surface complicates characterization of surface hydrophobicity. D.

Steric Forces

Steric forces are described by repulsive potentials that arise from near overlap between different moieties, be they atoms, molecules, or parts of macromolecules (e.g., segments or chains). They can manifest themselves in various forms for these different moieties [62], the most obvious being the overlap of the outer electron clouds of pairs of atoms as they approach each other, often referred to as Born repulsion. This situation (and the more complex one involving the approach of colloidal particles) can be modeled by the simple and intuitive hard-sphere potential, according to which there is no interaction until contact is reached, at which point the interaction potential jumps to infinity. A more accurate approach makes use of a slightly ‘‘softer’’ potential, for example, with an exponential or a power-law dependence, such as the Born repulsion term in the Lennard–Jones potential [Eq. (5)]. Steric repulsion can occur in several other forms and for more complex reasons [62]. Some of these phenomena owe their existence to the presence of solvent and may, thus, also be categorized as solvation forces. However, for proteins and other macromolecules the thermal fluctuations of surface groups may also play a role, with their suppression as solutes approach each other (or a surface) giving rise to an entropic repulsive force. The significance for proteins of contributions such as this one will depend on the mobility of surface chains and groups, and hence on the stability of the folded structure, but quantitative characterization remains problematic.

III.

RELATION TO BULK ADSORPTION MEASUREMENTS

While a number of insights regarding adsorption behavior have been gained from analysis of the configuration-dependent interaction energies discussed in the previous section, the energies alone are insufficient to make quantitative predictions regarding adsorption equilibrium, kinetics or adsorbed layer structure. Here we address the incorporation of molecular interactions into frameworks that allow such predictions to be formed and evaluated. Equilibrium extents of adsorption are most amenable to computation, whereas nonequilibrium behavior, including kinetic aspects and irreversible adsorption, is less so. Limitations in the quantitative understanding of the effects involved and in computational capabilities preclude full prediction of adsorption behavior in many © 2003 by Marcel Dekker, Inc.


83

situations. For example, molecular dynamics calculations applied to proteins at surfaces [8,9] can predict conformational changes within the constraints of their respective assumptions, but the configurational exploration required to predict adsorption equilibrium is not yet feasible. Thus the discussion below is limited to a relatively restricted set of conditions. In particular, the adsorbing molecules are assumed to be rigid and to retain their conformation during adsorption, an assumption that is likely to be satisfied mainly for ‘‘hard’’ globular proteins adsorbing on surfaces of limited hydrophobicity. Cases that are notable for not satisfying these assumptions include many widely studied experimental systems involving blood proteins such as albumin and fibrinogen adsorbing on polymetric biomaterials. On the other hand, systems in which long-range electrostatic forces are dominant are most amenable to mechanistic modeling, despite the fact that the adsorbed protein layers are expected to neutralize the adsorbent surface (e.g., lysozyme on mica [63]) at least partially. Here we consider first protein–surface interactions, addressing primarily the issues of adsorption reversibility and bulk–surface equilibrium. We then expand the discussion to include protein–protein interactions, which dictate ultimate surface coverage and adsorbed layer structure. A.

Protein–Surface Interactions

An analysis of protein–surface interactions is most usefully carried out at low surface coverages, where the confounding effect of interactions among adsorbate molecules is absent. The analysis is then usually performed in terms of equilibrium thermodynamics, specifically an adsorption equilibrium constant Keq = Cs /Cb, where Cs and Cb denote surface and bulk concentrations, respectively; since these concentrations are expressed per unit area and volume, respectively, Keq as defined here has units of length. Thus, Keq is a Henry’s law [64] constant, i.e., one in which the surface concentration is low enough that the adsorbed protein may be considered to be at infinite ‘‘dilution.’’ The link between the experimentally accessible Keq and the model calculations described previously is through the free energy of interaction, which is also often a quantity more easily interpretable in physical terms: the meaning of an adsorption equilibrium constant of 100 nm might not be clear, but a binding energy of 100 kJ/mol is likely to be more so. The most common expression relating free energy to the equilibrium constant is ⌬G = ⫺RT ln Keq

(8)

where ⌬G is the Gibbs free energy change between adsorbed and free protein. Since Keq is not dimensionless, ⌬G should in principle be expressed relative to that for a standard state, but in this case there is no clear way to define a standard state. The difficulty can be concealed when adjustable parameters are included in the freeenergy model. It is also possible to scale Keq by a characteristic length, e.g., a specific surface area (phase ratio) as used in chromatographic analysis, but incorporation of the geometric factor in the thermodynamic analysis is inconsistent with the dependence of adsorption equilibrium on the bulk concentration, rather than amount, of protein. An alternative approach that both reconciles these difficulties and probably better reflects the physical situation is one that abandons the notion of a single © 2003 by Marcel Dekker, Inc.

84

Roth and Lenhoff

adsorbed state in favor of a distribution of states in which protein is localized in the vicinity of the sorbent as a consequence of their favorable interaction. The extent to which the protein is localized depends directly on the free energy for a particular configuration (i.e., separation distance and mutual orientation) of protein and sorbent; specifically, the protein concentration should follow a Boltzmann distribution with respect to interaction energy. This idea of localization is essentially the Gibbs surface excess notion of adsorption and has been applied previously to gas adsorption [65]. The adsorbed amount is given as the total surface excess, the number of moles of solute actually present relative to that in the absence of the wall. This leads to the equilibrium constant, which for a homogeneous adsorbing surface is written as

冕冕 x

Keq =

dz

z0

(e⫺⌬F(⍀, z)/kT ⫺ 1) d⍀

(9)

⍀

where z is the solute–interface separation distance, z 0 is a cutoff distance, and ⍀ refers to orientational space, which must be included for nonspherically symmetric solute molecules such as proteins. Within this formalism, the equilibrium constant can be estimated from any model for the configuration-dependent total interaction energy ⌬F(⍀, z), which is typically computed as a Helmholtz free energy but is equivalent to the Gibbs free energy, as volume changes in the adsorption process are negligible. The assumption of equilibrium on which the discussion here has been based implies that adsorption is reversible. In many systems adsorption appears to be irreversible, but this may be a function of the experimental observation time. Typically, adsorption is diffusion limited with a characteristic time much shorter than the time required for desorption. Measurement of protein–surface equilibrium in nominally irreversible adsorption at low coverage may in fact be possible with long time measurements at extremely low bulk protein concentrations, where the mass transfer limitation is increased and as a result the rate of adsorption more closely matches the low desorption rate. Models of protein–surface equilibrium have been applied to a wide variety of adsorption data, even for high coverages. The amount of detail in such models has also varied, but in view of the uncertainties associated with several of the contributions discussed in Section II, more elaborate descriptions are not necessarily more successful at providing predictive mechanistic descriptions. They are, however, often capable of describing experimental trends with the aid of adjustable parameters, and in such situations an extensive data set, obtained under a range of conditions, is essential to test the model adequately. An example of such an approach is the model of Norde and Lyklema [66,67], in which the protein, adsorbate, solvent, and electrolyte are incorporated, and the effects accounted for include electrostatics, hydration, structural rearrangements, and transfer of hydrogen and other ions. However, there are uncertainties in the description of each mechanism. Models with a narrower focus allow less latitude, but obviously at the price of neglecting some effects. The most widely used have been those applied to electrostatically dominated systems, especially for correlating ion exchange chromatographic data, where adsorption must be reversible for operation to be successful. The earliest such model was a chemical one, the stoichiometric displacement model © 2003 by Marcel Dekker, Inc.


85

(SDM) of Boardman and Partridge [68], in which adsorption occurred as a protein displaced one or more ions on the sorbent in a strict ion exchange. The basic idea has been revised to include the concentration of ions in solution and the concept of distinct binding sites on the protein [69,70], but has remained quite popular for the correlation of ion exchange data. The essential result of this model is a linear dependence of log Keq on the logarithm of ionic strength, with the (negative) slope given by the net charge or number of binding sites on the protein, depending on interpretation. In either case, the slope parameter does not depend upon sorbent properties and so is not useful for scale-up or even the prediction of elution order. The SDM has been extended to account for behavior at high ionic strengths, where adsorption increases for reasons that are thought to be related to sequestration by electrolyte of the water molecules necessary to hydrate the protein. Solvophobic theory, based on the formation of a cavity to accept a solute, is able to correlate this salting out of protein [71,72], but again predictive capabilities are limited. Colloidal models are more amenable to a priori specification of at least protein properties such as size and charge, and in some cases to adsorbent properties as well. The application of such a model to describing retention in ion exchange chromatography was initially based on the colloidal interaction of two plates, for which the LPBE has an analytical solution [73,74]. Despite its geometric simplicity, this model successfully correlates ion exchange data both at low ionic strengths, where the retention is dominated by electrostatic attraction, and at higher ionic strengths. Two other interesting points emerged from this work. First, although log Keq is predicted by the model to be proportional to the inverse square root of ionic strength [46,47,74], as opposed to the inverse first power of ionic strength predicted by the SDM, the model adequately fitted data previously shown [70] to be described well by the SDM. Second, within the context of this model, the increase in retention at high ionic strengths is predicted to result from the screening of electrostatic repulsion that is due to oppositely charged, but greatly mismatched, surfaces. This is very different from the solvophobic mechanism previously discussed [70–72], but requires fitting several adjustable parameters, including unrealistically high surface charge densities. A higher level of realism in colloidal modeling is that in which the protein molecule is described as a sphere of radius R and net charge Q [37]. Electrostatic and van der Walls interaction energies are calculated as described in Sections II.A and II.B, and an equilibrium constant is obtained as in Eq. (9). The adsorbent, too, is characterized in terms of physically meaningful quantitative properties, namely, the surface charge density ␴ and the Hamaker constant A characterizing the material propensity for van der Walls interactions with proteins. In general, the model predicts that the charges on the protein and on the surface determine the steepness of the change in equilibrium constant with increasing ionic strength, whereas the Hamaker constant affects the level of the equilibrium constant [37]. The great benefit of this approach is that most of the parameter values can be specified without allowing them to be adjustable. The size of a protein molecule is estimated easily from its molecular weight. The charge on a protein or surface can be obtained from titration experiments, although this may be difficult for the sorbent if its specific area is low. For a protein, in fact, the primary sequence is usually sufficient to estimate the charge, as the titration curve is often not significantly affected by its local environment such as ionic strength [75]. The Hamakar constant © 2003 by Marcel Dekker, Inc.

86

Roth and Lenhoff

is the most difficult parameter to evaluate, but the Lifshitz theory provides a means to obtain at least an estimate, and Hamaker constants derived from calculation and experiment are available for a number of systems involving proteins [48,76–79]. Such models can be extended from a description of the protein as a sphere to one accounting more realistically for the geometry, based on crystallographic information [80] or on assumed confirmational forms [81]. The extension is, of course, more straightforward conceptually than it is in practice, but this barrier should diminish as computational capabilities continue to improve. Several of these modeling efforts have been used directly to compare with adsorption data obtained by ion exchange chromatography [82], fluorescence spectroscopy [80], and solution depletion [81]. Comparisons of calculations with adsorption trends, e.g., effects of mutations, have also been described [10,33]. Trends in Keq of multiple proteins and peptides with respect to ionic strength, and to a fair extent the quantitative values are adequately described by the models. Perhaps most remarkable, however, is how well grossly simplified models are able to describe the same adsorption behavior, e.g., with the protein or peptide represented as a point charge [81], plate [46,47,74], or sphere [80,82]. This robustness is probably a reflection of the fact that even simpler models can capture the dominant effect when electrostatics are globally attractive, with an implicit or explicit adjustable parameter aiding in ‘‘tuning’’ the actual values of adsorption constants. Trends such as those with varying ionic strength are then more easily reproduced. When the adsorption behavior is qualitatively less easily predictable, e.g., when electrostatics are generally repulsive but patch-controlled adsorption occurs, results are more sensitive to modeling details [35]. Although the adsorbent physicochemical properties, namely, the surface charge density and the Hamaker constant for interactions with proteins, are well-defined physical quantities, there is uncertainty in how meaningful they are in characterizing protein–surface interactions. Among the sources of uncertainty are the effects of discreteness of charge, the adsorbent geometry (to which van der Waals interactions are especially sensitive [48]), and neglect of other effects such as hydration-related ones. Thus the adsorbent properties included in the model may be surrogates for other effects in addition to those that they are nominally intended to capture. In particular, descriptions of van der Waals interactions, being of short range, may also account to some extent for other short-range interactions such as solvation interactions; these have in common their strong dependence on the size of the interacting areas on the protein and surface. This view gives rise to a simpler but more pragmatic formulation in which the protein electrostatics are represented by a sphere with net charge, while all other effects are captured in a short-range interaction energy that increases with protein size [82]. B.

Protein–Protein Interactions

At higher coverages than those considered in Section III.A, the Henry’s law limit implied by the use of an equilibrium constant Keq is no longer observed, and the finite amount of surface area available comes into play in the absence of multilayer adsorption. The result is then that Cs no longer increases linearly with Cb , but instead that a plot of Cs versus Cb typically adopts the convex-upward shape characteristic of adsorption isotherms of many kinds. © 2003 by Marcel Dekker, Inc.


87

Adsorption behavior of this kind is most frequently described using a Langmuir adsorption isotherm, despite the widespread recognition that most of the assumptions underlying this model are not satisfied for protein adsorption. The Langmuir formulation assumes adsorption to occur on discrete sites, to be reversible, and interactions among sites to be negligible, and the hyperbolic form of the isotherm is then a consequence of the finite number of sites available. For adsorption of proteins, all of these assumptions become questionable. First, the issue of reversibility has been mentioned earlier and is discussed again subsequently. Next, the adsorbate molecules are so much larger than the surface lattice dimensions that the adsorbent is more reasonably considered to be a continuum than comprising discrete sites. Thus, the finite number of sites in the Langmuir model should be replaced by consideration of the finite adsorbent area. More generally, this excluded-area feature may be thought of in terms of steric interactions among adsorbate molecules, and in this context it is also necessary to consider protein–protein energetic interactions on the surface. The most obvious energetic interactions are long-range electrostatics, which will generally be repulsive in view of the like charge carried by the protein molecules. Shorter-range interactions become more important as complete monolayer coverage is approached, but their effects are more difficult to assess. Because such attractive contributions as van der Waals and hydrophobic interactions are strongly dependent on molecular complementarity [48], strong attraction is not as likely as would be expected for general colloidal particles at short range. Beyond just the energetics of the interactions is also the possibility, even for conformationally rigid molecules, of orientational adaptations as the coverage increases [31,83]. Because of these and other effects, the fits of the Langmuir equation to protein adsorption data often show systematic discrepancies. In particular, experimental isotherms are often ‘‘softer’’ than can be fitted by the Langmuir form, i.e., the approach to a plateau is less steep than predicted, and the plateau region may show a continued gradual increase in Cs , with coverage still not exceeding theoretical monolayer levels as estimated from simple geometrical packing arguments. Such experimental isotherms can be adequately described by various alternative models. For instance, the steric mass action (SMA) model [84] is an extension of the SDM discussed earlier for ion exchange chromatography, and although it is not explicitly formulated in terms of excluded area, the essential features are analogous. The resulting isotherms are thus the convex-upward ones widely observed in practice. For more mechanistically based models of high-coverage adsorption, however, the formulation and solution are complicated by the many-body nature of the problem, specifically the need to know the relative positions of the molecules involved, i.e., the arrangement of molecules on the surface. The configuration directly affects the protein–protein energetics, and there is generally a trade-off between this, generally repulsive, contribution, and the attractive protein–surface energetics. In addition, the questions raised earlier of whether adsorption is reversible and whether the adlayer is an equilibrium phase must be considered. To illustrate some of the considerations involved, we examine various examples of irreversible adsorption first. Several possibilities arise even in the absence of protein–protein energetic interactions, as can be seen by exploring the predicted adsorption behavior when the proteins are modeled as simple hard spheres. From simple geometric arguments, © 2003 by Marcel Dekker, Inc.

88

Roth and Lenhoff

close-packed hexagonal and square layers have coverages of 90.7 and 78.5% respectively. However, such close packing may be unattainable. For the limiting case in which the spheres approach the surface sequentially and are adsorbed irreversibly, if there is no overlap with previously adsorbed spheres, the process is known as random sequential adsorption (RSA), and the maximum coverage, the so-called jamming limit, is about 54.7% [85,86]. If the RSA model is modified to allow sterically excluded particles to ‘‘roll over’’ the blocking particles in order to reach the surface (ballistic deposition), the ultimate coverage is about 61% [87,88]. Thus, the randomness of the locations at which particles adsorb results in a considerable additional fraction of uncovered surface. When there are energetic interactions, especially repulsive electrostatics, between the protein molecules, the picture becomes more complicated. The RSA model has been modified to include an energetic threshold for rejection of a potential adsorption event [89–91], leading to jamming limits lower than the usual 54.7%. Although this model captures observed experimental behavior, a more likely explanation of the physics involved is that repulsion by adsorbed molecules presents a kinetic barrier precluding additional adsorption [91]. In either case, an increase in salt concentration reduces protein–protein repulsion, leading to a predicted increase in adsorbed amount. For relatively small globular proteins, however, the repulsion may be too weak for the modified RSA models to be valid. The stated bases for estimating surface coverage are premised on the assumption that adsorption is irreversible. One consequence of this is that the coverage is independent of the bulk protein concentration, i.e., any ‘‘isotherm’’ would rise infinitely steeply to the plateau. The issue of reversibility is also paramount if adsorbed particles are able to diffuse on the surface: irreversible adsorption would ultimately lead to a close-packed surface layer. Thus, if particles are able to move on the surface, it is reversible adsorption that is usually of interest, with the limiting behavior at low coverage represented by the equilibrium constant Keq as discussed in the previous section. Configurational issues are crucial here; statistical mechanical approaches are possible for simple intermolecular potentials [92], but simulations are required for more realistic representations. As the coverage increases, the energy gain due to adsorption (protein–surface interaction) becomes offset by a penalty due to protein–protein interaction. This penalty is present even for hard spheres devoid of energetic interactions with one another, which give rise to a surface pressure if they are mobile. Such interactions can be accounted for by treating the adsorbate layer as a two-dimensional fluid [93]. As Fig. 5 shows, the resulting isotherm is appreciably ‘‘softer’’ than its Langmuir counterpart; the figure also shows, for comparison, the coverages predicted by the various models invoking irreversible adsorption. That the abscissa is in terms of KeqCb illustrates nicely the central importance of the protein–surface interactions, offset by the increasing significance of protein–protein interactions as the coverage increases. However, the form of the abscissa somewhat obscures some important features of the predicted isotherms. First, the discrepancies between the Langmuir and the softer isotherm are exaggerated by the logarithmic axis. Second, scaling by Keq can result in conspicuous changes in isotherms plotted on the more conventional linear abscissa Cb. Specifically, for small Keq the isotherms rise rather gradually, and a clearly identifiable plateau is not seen, whereas for high Keq a steeper (high-affinity) isotherm and fairly flat plateau result. Thus for electrostatically driven adsorption, © 2003 by Marcel Dekker, Inc.


89

FIG. 5 Fractional surface coverages predicted by various models of adsorption of spheres.

the reduction in Keq with increasing salt concentration is predicted to give rise to a monotonic decrease in the adsorbed amount at a given protein concentration. This is in fact what is usually seen under conditions such as those encountered in ion exchange chromatography, but it is opposite to the trend described here resulting from screening of protein–protein interactions, e.g., in RSA. The ideal array concept can also be modified to include energetic interactions in describing equilibrium adsorption [94]. Accounting for electrostatic and van der Waals interactions leads to isotherms with a generally ‘‘soft’’ convex-upward form, with the same trend with increasing salt as described in the previous paragraph, i.e., screening of protein–surface interactions is more important than screening of protein–protein interactions. However, a direct comparison with Fig. 5 is difficult because of the different scaling involved. Beyond the general isotherm form, the existence of multiple-valued isotherms was predicted. This result, which is qualitatively similar to previous predictions for different models of adsorption, suggests possible explanations for apparent anomalies in protein adsorption isotherms, e.g., steps and kinks. The general case in which reversibility, mobility, and randomness are all possible can be undertaken only by simulation methods such as Brownian dynamics [95,96], with tractability generally requiring use of pairwise additive analytical approximations for evaluating potentials. The grand canonical Brownian dynamics scheme [96] accounts directly for all interactions, configurational issues, and adsorption dynamics. For a small globular protein such as lysozyme at pH 7, it predicts a range of adsorption behaviors under electrostatically controlled conditions. At very low salt concentrations, the predicted isotherms show high-affinity behavior, with © 2003 by Marcel Dekker, Inc.

90

Roth and Lenhoff

the plateau amounts increasing with ionic strength; i.e., repulsive protein–protein interactions dominate. At higher ionic strengths, however, the trend predicted by Chatelier and Minton [93] and Johnson et al. [94] is recovered. The number of degrees of freedom in the case of high surface coverage leaves this as an extremely complex situation, even when the individual molecules are treated in an idealized fashion. Allowing for the anisotropy of protein molecules adds complexity of a different kind, possibly leading to counterintuitive behavior, e.g., favorable intermolecular electrostatic interactions even for molecules with a nonzero net charge [32]. Even simple packing arguments are complicated by the different sizes of the protein ‘‘footprint’’ in different orientations.

IV.

SUMMARY

The primary thrust of this chapter is the quantitative determination of protein–surface interaction energies, and the implications of such information for experimentally accessible quantities, e.g., adsorption isotherms. The calculation of interaction energies is feasible for systems dominated by electrostatic and van der Waals interactions, with both rigorous approaches and simpler approximations available. Such calculations are, however, predicated on the availability of protein structural information and surface characteristics, as well as on the assumption that both the protein and the surface are fairly rigid. When these requirements are satisfied, adsorption equilibrium can be predicted quite well in absolute terms, and key trends are captured very well. In particular, these calculations provide a meaningful solid fundamental framework within which to develop rational methods for manipulating adsorption. Although these capabilities are valuable, many challenges remain, and in view of the complexity, even qualitative, that is involved with some of them, it is not clear what solutions may emerge. Probably the most important outstanding features are characterization and modeling of hydrophobic interactions and understanding the cause and effect of protein conformational changes during adsorption.

REFERENCES 1. 2. 3. 4. 5. 6. 7.

J. D. Andrade and V. Hlady, Protein adsorption and materials biocompatability: a tutorial review and suggested hypotheses. Adv. Polym. Sci. 79:1–63 (1986). W. Norde, Adsorption of proteins from solution at the solid–liquid interface. Adv. Coll. Interf. Sci. 25:267–340 (1986). C. A. Haynes and W. Norde, Globular proteins at solid/liquid interfaces. Colloids Surf. B: Biointerfaces 2:517–566 (1994). W. Norde and J. Lyklema, Why proteins prefer interfaces. J. Biomater. Sci. Polym. Edn. 2:183–202 (1991). D. R. Lu and K. Park, Protein adsorption on polymer surfaces: calculation of adsorption energies. J. Biomater. Sci. Edn. 1:243–260 (1990). D. R. Lu, S. J. Lee, and K. Park, Calculation of solvation interaction energies for protein adsorption onto polymer surfaces. J. Biomater. Sci. Polym. Edn. 3:127–147 (1990). S. J. Lee and K. Park, Protein interactions with surfaces: separation distance-dependent interaction energies. J. Vac. Sci. Technol. A 12:2949–2955 (1994).


Quantitative Modeling of Protein Adsorption 8.

9. 10.

11.

12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

24. 25.

26. 27. 28.

29. 30.

91

K. Lim and J. N. Herron, Molecular simulation of protein–PEG interaction. In Poly (Ethylene Glycol) Chemistry: Biotechnical and Biomedical Applications ( J. M. Harris, ed.). Plenum Press, New York, 1992, pp. 29–57. D. J. Tobias, W. Mar, J. K. Blasie, and M. L. Klein, Molecular dynamics simulations of a protein on hydrophobic and hydrophilic surfaces. Biophys. J. 71:2933–2941 (1996). V. Noinville, C. Vidal-Madjar, and B. Se´bille, Modeling of protein adsorption on polymer surfaces. Computation of adsorption potential. J. Phys. Chem. 99:1516–1522 (1995). S. J. Weiner, P. A. Kollman, D. A. Case, U. C. Singh, C. Ghio, G. Alagona, S. Profeta, and P. Weiner, A new force field for molecular mechanical simulation of nucleic acids and proteins. J. Am. Chem. Soc. 106:765–784 (1984). I. Szleifer, Protein adsorption on surfaces with grafted polymers: a theoretical approach. Biophys. J. 72:595–612 (1997). T. Arai and W. Norde, The behavior of some model proteins at solid–liquid interfaces. I. Adsorption from single protein solution. Colloids Surf. 51:1–15 (1990). J. G. Kirkwood, Theory of solutions of molecules containing widely separated charges with special application to zwitterions. J. Chem. Phys. 2:351–361 (1934). R. Hogg, T. W. Healy, and D. W. Fuerstenau, Mutual coagulation of colloidal dispersions. Trans. Faraday Soc. 62:1638–1651 (1966). J. T. G. Overbeek, The role of energy and entropy in the electrical double layer. Colloids Surf. 51:61–75 (1990). S. A. Palkar and A. M. Lenhoff, Energetic and entropic contributions to the interaction of unequal spherical double layers. J. Coll. Interf. Sci. 165:177–194 (1994). H.-X. Zhou, Macromolecular electrostatic interaction energy within the nonlinear Poisson–Boltzmann equation. J. Chem. Phys. 100:3152–3162 (1994). K. A. Sharp and B. Honig, Electrostatic interactions in macromolecules: theory and applications. Ann. Rev. Biophys. Chem. 19:301–332 (1990). B. J. Yoon and A. M. Lenhoff, Computation of the electrostatic interaction energy between a protein and a charged surface. J. Phys. Chem. 96:3130–3134 (1992). M. K. Gilson, K. A. Sharp, and B. Honig, Calculating the electrostatic potential of molecules in solution: method and error assessment. J. Comp. Chem. 9:327–335 (1988). J. Warwicker and H. C. Watson, Calculation of the electric potential in the active site cleft due to ␣-helix dipoles. J. Mol. Biol. 157:671–679 (1982). W. H. Orttung, Direct solution of the Poisson equation for biomolecules of arbitrary shape, polarizability density, and charge distribution. Ann. NY Acad. Sci. 303:22–37 (1977). R. J. Zauhar and R. S. Morgan, A new method for computing the macromolecular electric potential. J. Mol. Biol. 186:815–820 (1985). A. H. Juffer, E. F. F. Botta, B. A. M. van Keulen, A. van der Plug, and J. C. Berendsen, The electric potential of a macromolecule in a solvent: a fundamental approach. J. Comp. Phys. 97:144–171 (1991). B. J. Yoon and A. M. Lenhoff, A boundary element method for molecular electrostatics with electrolyte effects. J. Comp. Chem. 11:1080–1086 (1990). H.-X. Zhou, Boundary element solution of macromolecular electrostatics: interaction energy between two proteins. Biophys. J. 65:955–963 (1993). D. Bashford and M. Karplus, Multiple site titration curves of proteins: an analysis of exact and approximate methods for their calculation. J. Phys. Chem. 95:9556–9561 (1991). M. K. Gilson and B. H. Honig, Energetics of charge–charge interactions in proteins. Proteins Struct. Funct. Genet. 3:32–52 (1988). L. Haggerty and A. M. Lenhoff, Relation of protein electrostatics computations to ionexchange and electrophoretic behavior. J. Phys. Chem. 95:1472–1477 (1991).


92

Roth and Lenhoff

31.

C. S. Lee and G. Belfort, Changing activity of ribonuclease A during adsorption: a molecular explanation. Proc. Natl. Acad. Sci. USA 86:8392–8396 (1989). L. Haggerty and A. M. Lenhoff, Analysis of ordered arrays of adsorbed lysozyme by scanning tunneling microscopy. Biophys. J. 64:886–895 (1993). D. J. Roush, D. S. Gill, and R. C. Willson, Electrostatic potentials and electrostatic interaction energies of rat cytochrome b5 and a simulated anion-exchange adsorbent surface. Biophys. J. 66:1290–1300 (1994). V. Lesins and E. Ruckenstein, Patch controlled attractive electrostatic interactions between similarly charged proteins and adsorbents. Colloid. Polym. Sci. 266:1187–1190 (1988). D. Asthagiri and A. M. Lenhoff, Influence of structural details in modelling electrostatically driven protein adsorption. Langmuir 13:6761–6768 (1997). D. J. Barlow and J. M. Thornton, The distribution of charged groups in proteins. Biopolymers 25:1717–1733 (1986). C. M. Roth and A. M. Lenhoff, Electrostatic and van der Waals contributions to protein adsorption: computation of equilibrium constants. Langmuir 9:962–972 (1993). M. L. Grant and D. A. Saville, Colloidal interactions in protein crystal growth. J. Phys. Chem. 98:10358–10367 (1994). C. M. Roth, Electrostatic and van der Waals contributions to protein adsorption. PhD dissertation, University of Delaware, Newark, DE, 1994. H. C. Hamaker, The London–van der Waals attraction between spherical particles. Physica 4:1058–1072 (1937). I. E. Dzyaloshinskii, E. M. Lifshitz, and L. P. Pitaevskii, The general theory of van der Waals forces. Adv. Phys. 10:165–209 (1961). D. J. Mitchell and B. W. Ninham, van der Waals forces between two spheres. J. Chem. Phys. 56:1117–1126 (1972). J. E. Kiefer, V. A. Parsegian, and G. H. Weiss, An easily calculable approximation for the many-body van der Waals attraction between two equal spheres. J. Coll. Interf. Sci. 57:580–582 (1976). J. Lyklema, Fundamentals of Interface and Colloid Science. Volume 1: Fundamentals. Academic Press, London, 1991. E. Ruckenstein and D. C. Prieve, Adsorption and desorption of particles and their chromatographic separation. AIChE J. 22:276–283 (1976). J. Sta˚hlberg, B. Jo¨nsson, and C. Horva´th, Combined effect of coulombic and van der Waals interactions in the chromatography of proteins. Anal. Chem. 64:3118–3124 (1992). J. Sta˚hlberg and B. Jo¨nsson, Influence of charge regulation in electrostatic interaction chromatography of proteins. Anal. Chem. 68:1536–1544 (1996). C. M. Roth, B. L. Neal, and A. M. Lenhoff, Van der Waals interactions involving proteins. Biophys. J. 70:977–987 (1996). D. B. Hough and L. R. White, The calculation of Hamaker constants from Lifshitz theory with applications to wetting phenomena. Adv. Coll. Interf. Sci. 14:3–41 (1980). V. A. Parsegian, Long range van der Waals forces. In Physical Chemistry: Enriching Topics from Colloid and Surface Science (H. van Olphen and K. J. Mysels, eds.). Theorex, La Jolla, CA, 1975, pp. 27–72. T. Inagaki, R. N. Hamm, E. T. Arakawa, and R. D. Birkhoff, Optical property of bovine plasma albumin between 2 and 82 eV. Biopolymers 14:839–845 (1975). S. Nir, van der Waals interactions between surfaces of biological interest. Progr. Surf. Sci. 8:1–58 (1976). J. Israelachvili, Solvation forces and liquid structure as probed by direct force measurements. Acc. Chem. Res. 20:415–421 (1987).

32. 33.

34.

35. 36. 37. 38. 39. 40. 41. 42. 43.

44. 45. 46.

47. 48. 49. 50.

51. 52. 53.


Quantitative Modeling of Protein Adsorption 54.

55. 56. 57. 58. 59. 60. 61. 62. 63.

64. 65. 66.

67.

68. 69. 70.

71.

72.

73. 74. 75.

76.

93

R. P. Rand, N. Fuller, V. A. Parsegian, and D. C. Rau, Variation in hydration forces between neutral phospholipid bilayers: evidence for hydration attraction. Biochemistry 27:7711–7722 (1988). R. D. Tilton, C. R. Robertson, and A. P. Gast, Manipulation of hydrophobic interactions in protein adsorption. Langmuir 7:2710–2718 (1991). A. Ben-Naim, Hydrophobic Interactions. Plenum Press, New York, 1990. C. Tanford, The Hydrophobic Effect. Wiley, New York, 1980. T. E. Creighton, Proteins: Structures and Molecular Properties. 2nd ed. Freeman, New York, 1993. D. Eisenberg and A. D. McLachlan, Solvation energy in protein folding and binding. Nature 319:199–203 (1986). D. J. Abraham and A. J. Leo, Extension of the fragment method to calculate amino acid zwitterion and side chain partition coefficients. Proteins 2:130–152 (1987). J. Israelachvili and R. Pashley, The hydrophobic interaction is long range, decaying exponentially with distance. Nature 300:341–342 (1982). J. N. Israelachvili, Intermolecular and Surface Forces. 2nd ed. Academic Press, New York, 1992. E. Blomberg, P. M. Claesson, J. C. Froberg, and R. D. Tilton, The interaction between adsorbed layers of lysozyme studied with the surface force technique. Langmuir 10: 2325–2334 (1994). S. I. Sandler, Chemical and Engineering Thermodynamics. 2nd ed. Wiley, New York, 1989. J. A. Barker and D. H. Everett, High temperature adsorption and the determination of the surface area of solids. Trans. Faraday Soc. 58:1608–1623 (1962). W. Norde and J. Lyklema, The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces: IV. The charge distribution in the adsorbed state. J. Coll. Interf. Sci. 66:285–294 (1978). W. Norde and J. Lyklema, Thermodynamics of protein adsorption: theory with special reference to the adsorption of human plasma albumin and bovine pancreas ribonuclease at polystyrene surfaces. J. Coll. Interf. Sci. 71:350–366 (1979). N. K. Boardman and S. M. Partridge, Separation of neutral proteins on ion-exchange resins. Biochem. J. 59:543–552 (1955). W. Kopaciewicz, M. A. Rounds, J. Fausnaugh, and F. E. Regnier, Retention model for high-performance ion exchange chromatography. J. Chromatogr. 266:3–21 (1983). W. R. Melander, Z. El Rassi, and C. Horva´th, Interplay of hydrophobic and electrostatic interactions in biopolymer chromatography: effect of salts on the retention of proteins. J. Chromatogr. 469:3–27 (1989). W. Melander and C. Horva´th, Salt effects on hydrophobic interactions in precipitation and chromatography of proteins: an interpretation of the lyotropic series. Arch. Biochem. Biophys. 183:200–215 (1977). W. R. Melander, D. Corradini, and C. Horva´th, Salt-mediated retention of proteins in hydrophobic-interaction chromatography: application of solvophobic theory. J. Chromatogr. 317:67–85 (1984). V. A. Parsegian and D. Gingell, On the electrostatic interaction across a salt solution between two bodies bearing unequal charges. Biophys. J. 12:1192–1204 (1972). J. Sta˚hlberg, B. Jo¨nsson, and C. Horva´th, Theory for electrostatic interaction chromatography of proteins. Anal. Chem. 63:1867–1874 (1991). Y. Nozaki and C. Tanford, Examination of titration behavior. In Methods in Enzymology. Vol. XI: Enzyme Structure (C. Hirs, ed.). Academic Press, New York, 1967, pp. 715– 734. V. A. Parsegian and S. L. Brenner, The role of long range forces in ordered arrays of tobacco mosaic virus. Nature 259:632–635 (1976).


94

Roth and Lenhoff

77.

T. Afshar-Rad, A. I. Bailey, P. F. Luckham, W. MacNaughton, and D. Chapman, Forces between protein and model polypeptides adsorbed on mica surfaces. Biochim. Biophys. Acta 915:101–111 (1987). C. A. Helm, W. Knoll, and J. N. Israelachvili, Measurements of ligand–receptor interactions. Proc. Natl. Acad. Sci. USA 88:8169–8173 (1991). D. E. Leckband, F.-J. Schmitt, J. N. Israelachvili, and W. Knoll, Direct measurements of specific and nonspecific protein interactions. Biochemistry 33:4611–4624 (1994). C. M. Roth and A. M. Lenhoff, Electrostatic and van der Waals contributions to protein adsorption: comparison of theory and experiment. Langmuir 11:3500–3509 (1995). N. Ben-Tal, B. Honig, R. M. Peitzsch, G. Denisov, and S. McLaughlin, Binding of small basic peptides to membranes containing acidic lipids: theoretical models and experimental results. Biophys. J. 71:561–575 (1996). C. M. Roth, K. K. Unger, and A. M. Lenhoff, A mechanistic model of retention in protein ion-exchange chromatography. J. Chromatogr. A 726:45–56 (1996). J. L. Robeson and R. D. Tilton, Spontaneous reconfiguration of adsorbed lysozyme layers observed by total internal reflection fluorescence with a pH-sensitive fluorophore. Langmuir 12:6104–6113 (1996). C. A. Brooks and S. M. Cramer, Steric mass-action ion exchange: displacement profiles and induced salt gradients. AIChE J. 38:1969–1978 (1992). J. Feder, Random sequential adsorption. J. Theor. Biol. 87:237–254 (1980). J. Feder and I. Giaever, Adsorption of ferritin. J. Coll. Interf. Sci. 78:144–154 (1980). R. Jullien and P. Meakin, Random sequential adsorption with restructuring in two dimensions. J. Phys. A: Math. Gen. 25:L189–L194 (1992). H. S. Choi, J. Talbot, G. Tarjus, and P. Viot, First-layer formation in ballistic deposition of spherical particles: kinetics and structure. J. Chem. Phys. 99:9296–9303 (1993). Z. Adamczyk, M. Zembala, B. Siwek, and P. Warszynski, Structure and ordering in localized adsorption of particles. J. Coll. Interf. Sci. 140:123–137 (1990). Z. Adamczyk, B. Siwek, M. Zembala, and P. Belouschek, Kinetics of localized adsorption of colloid particles. Adv. Coll. Interf. Sci. 48:151–280 (1994). M. R. Oberholzer, J. M. Stankovich, S. L. Carnie, D. Y. Chan, and A. M. Lenoff, 2-D and 3-D interactions in random sequential adsorption of charged particles. J. Coll. Interf. Sci. 194:138–153 (1997). D. Nicholson and N. G. Parsonage, Computer Simulation and the Statistical Mechanics of Adsorption. Academic Press, London, 1982. R. C. Chatelier and A. P. Minton, Adsorption of globular proteins on locally planar surfaces: models for the effect of excluded surface area and aggregation of adsorbed protein on adsorption equilibria. Biophys. J. 71:2367–2374 (1996). C. A. Johnson, P. Wu, and A. M. Lenhoff, Electrostatic and van der Waals contributions to protein adsorption: 2. Modelling of ordered arrays. Langmuir 10:3705–3713 (1994). E. Dickinson, Brownian dynamics with hydrodynamic interactions: the application to protein diffusional problems. Chem. Soc. Rev. 14:421–455 (1985). M. R. Oberholzer, N. J. Wagner, and A. M. Lenhoff, Grand canonical Brownian dynamics simulation of colloidal adsorption. J. Chem. Phys. 107:9157–9167 (1997). J. E. Sader and A. M. Lenoff, J. Coll. Interf. Sci. (1998).

78. 79. 80. 81.

82. 83.

84. 85. 86. 87. 88. 89. 90. 91.

92. 93.

94. 95. 96. 97.


5 Interfacial Behavior of Protein Mutants and Variants MARTIN MALMSTEN Institute for Surface Chemistry and Royal Institute of Technology, Stockholm, Sweden THOMAS ARNEBRANT Institute for Surface Chemistry, Stockholm, and Malmo¨ University, Malmo¨, Sweden PETER BILLSTEN*

I.

Linko¨ping University, Linko¨ping, Sweden

INTRODUCTION

The interfacial behavior of proteins is of importance in biochemical and biophysical processes, such as complement activation, thrombus formation, initial stages of arteriosclerosis, dental pellicle formation, etc., and a host of biomedical applications, such as intravenous drug delivery, solid-phase diagnostics, extracorporeal therapy, biomaterials, biosensors, biotechnical separation methods, biofouling, dental implants, etc. Not surprisingly, therefore, substantial work has been devoted to understanding the behavior of proteins at interfaces and the interplay between factors determining this interfacial behavior. Despite considerable progress in the understanding of these processes, regarding, e.g., the relative importance of electrostatic effects, conformational stability, etc., and of adsorption kinetics and interfacial exchange phenomena, an undeniable fact is that the current understanding of protein interfacial behavior lags that of the interfacial behavior of homopolymers, copolymers, polyelectrolytes, and polyampholytes [1–3]. Naturally, this is a consequence of the higher complexity of proteins as compared to these simpler macromolecules. However, it is also possible that the lack of awareness among scientists concerned with protein adsorption about recent progress in the field of polymer adsorption is partly to blame for the current situation. Although it is important to remember that proteins are more complex than simpler polymers, they are still subject to the same thermodynamics, and therefore at least some general features should be analogous to those displayed by simpler systems. Indeed, this view is stressed repeatedly throughout this book. One of the complexities of proteins is their rather unique structure, which to some extent has hindered systematic investigation of the effects of e.g., molecular weight, protein–surface interactions, structural stability, etc., on the protein adsorption, since more often than not such investigations have involved the comparison of proteins differing in more than one respect. For example, when investigating the *Current affiliation: Analysis and Formulations, Astra Draco AB, Lund, Sweden.


96

Malmsten et al.

effect of the molecular weight, typically totally different proteins have been compared, which means that not only the molecular weight, but also, e.g., the protein structure and structural stability, as well as the protein isoelectric point, net charge, and charge distribution are different, precluding a conclusive analysis. Although there are some naturally occurring proteins existing in several quite similar variants which can be, and have been, used for this type of investigation, it is mainly after the recent developments in protein engineering that they have really become feasible. It is the aim of the present chapter to try to exemplify how investigations with both naturally occurring protein variants and protein mutants may yield information on the effects of protein molecular weight, structural stability, and protein–surface and protein–protein interactions on the interfacial behavior of proteins. Note, however, that it is not our aim to make a complete inventory of reported studies with protein variants and mutants. For convenience, we have subdivided the different types of mutants (variants) rather arbitrarily and nonstringently into three categories —stability mutants, (surface) interaction mutants, and association mutants, the name indicating the main effect of the mutation (variation). Where appropriate, comparisons will be made between the findings for the protein mutants and variants on one hand and those for simpler macromolecules, e.g., homopolymers, copolymers, and polyelectrolytes, on the other.

II.

INTERFACIAL BEHAVIOR OF NATURALLY OCCURRING PROTEIN VARIANTS

One natural starting point for systematic studies of the effects of, e.g., molecular weight, conformational stability, and protein – surface and protein – protein interactions for the interfacial behavior of proteins is to investigate the adsorption properties of naturally occurring variants of the same protein, and there are several papers where this approach has been used. For example, in an early study Horseley et al. investigated the adsorption of hen egg white and human milk lysozyme at various surfaces with total internal reflection fluorescence spectroscopy (TIRF), and found that these proteins indeed behave differently at interfaces [4]. A more recent study by Xu and Damodaran with those and other variants has shown that these differences exist also at the air – water interface [5]. The latter investigation also clearly indicated that the adsorption at this interface increases with the degree of protein denaturation (see subsequent discussion). Furthermore, as will be discussed in more detail, the ␤-lactoglobulin variants A and B, displaying different selfassembly behavior, were investigated by Elofsson et al. [6 – 9]. Yet another example of a study of the interfacial behavior of naturally occurring protein variants is that by Elbaum et al., in which the surface activity of hemoglobin S and other human hemoglobin variants was investigated, and in which the variants hemoglobin S and hemoglobin CHarlem, both containing ␤6Glu → Val substitutions, were found to be more surface active than hemoglobin A, hemoglobin C, and hemoglobin Korle Bu (␤73Asp → Asn) [10]. This was interpreted as being related to the ligated state of the ␤ s chain in the tetramer. However, as for the other examples given above, with the exception of the ␤-lactoglobulin variants in the studies by Elofsson et al., the mechanistic interpretation of the differences in interfacial behavior of these variants is somewhat difficult. © 2003 by Marcel Dekker, Inc.

Protein Mutants and Variants

III.

97

INTERFACIAL BEHAVIOR OF STABILITY MUTANTS

It has been found for a range of proteins that their tendency for adsorption at various surfaces increases on approaching their thermal denaturation. For example, this behavior has been observed for a series of lysozyme variants by Xu and Damodaran [5] and for ␤-lactoglobulin by Elofsson et al. [9] and Arnebrant et al. [11]. Analogously, proteins capable of undergoing interfacial conformational change are frequently found to adsorb at surfaces even when electrostatics counteracts adsorption [2,3]. There could, in principle, be several reasons for this. One of these could be that on undergoing conformational changes amino acids typically hidden in the protein interior may be made accessible to interaction with both the surface and the solvent. Since these residues typically are hydrophobic, the overall solvency of the protein could decrease. This could be expected to lead to an increased surface activity due to both direct solubility effects and indirect effects due to solvency-induced protein self-assembly, analogous to the performance of simpler systems, such as homo- and copolymers [1], and to the maximum adsorption frequently found for proteins at their isoelectric point [2,3]. Furthermore, it has been observed that the adsorption of proteins at various interfaces may be entropically driven, and it has been suggested that one origin of this increase in the entropy is the loss of ordered structure, e.g., ␣-helices, on adsorption, which in fact has been observed in numerous experimental systems [2,3] (see following discussion). However, other mechanisms for the observed entropy gain on adsorption can also be envisaged, including, e.g., the release of hydration water and/or counterions for both the protein and the surface on adsorption. Protein mutants offer interesting possibilities for the investigation of the effects of the conformational stability on the adsorption of proteins, since structure mutants are fairly straightforwardly achievable by protein engineering at an essentially fixed molecular weight and unchanged composition of residues at the protein surface [12]. Not surprisingly, therefore, structural stability investigations constitute one of the most frequent of the different types of studies possible with protein mutants. Until today several different systems have been investigated, including T4 lysozyme [12– 17], tryptophane synthase ␣-subunits [18], and human carbonic anhydrase II [19,20]. One of the first systems investigated regarding the effects of protein structural stability on its interfacial behavior was tryptophan synthase ␣-subunits, which is available in numerous different mutants ranging in denaturation free energy between about 5 and 10 kcal/mol (at pH 9), which can be further controlled by varying pH. As found by Kato and Yutani, these have very different surface activity [18]. More precisely, a clear correlation was found between the protein structural stability and its surface activity, since the air–water surface tension was found to vary in an essentially linear manner with the free energy of denaturation. Since the surface and interfacial tension are crucial parameters for foam formation and emulsification, also the latter were found to depend on the protein structural stability in a similar manner —i.e., the less stable the protein, the better the emulsification and the better the foam stability. However, only a limited amount of information on the mechanisms of the increased interfacial activity with a decreasing protein structural stability was obtained from these experiments. Analogous to the tryptophan synthase ␣-subunits, bacteriophage T4 lysozyme is available in a range of conformational stabilities, achieved primarily through point mutations in the Ile3 position [12]. Also the interfacial behavior of these mutants © 2003 by Marcel Dekker, Inc.

98

Malmsten et al.

has been investigated. For example, McGuire et al. investigated the adsorption of a series of T4 lysozyme stability mutants at silica and methylated silica surfaces with ellipsometry, and found some trends for the equilibrium adsorbed amount at both these surfaces to follow the protein structural stability (Table 1) [13]. A somewhat more straightforward relationship appeared to exist between the degree of removal (‘‘elutability’’) of the adsorbed protein by a cationic surfactant and the structural stability, indicating the less stable proteins undergo larger interfacial conformational changes. (Subsequent investigations showed that removal by an ionic surfactant occurred at a lower surfactant concentration for the less stable mutant than for the wildtype protein. From the ionic strength dependence it was inferred that this was due to an earlier onset of the cooperative binding of the surfactant to the less stable, i.e., more structurally altered, protein [39].) A similar conclusion was reached from an analysis of the adsorption kinetics displayed by these mutants [17]. Later, this was shown directly by Billsten et al. for T4 lysozyme at silica, using circular dichroism (CD) [15]. As seen in Fig. 1, the helical content after adsorption at silica of three proteins with different structural stability but with essentially identical helical content in solution decreases with the protein structural stability. Note that the kinetics of interfacial conformational change are also affected by the protein structural stability and increase with decreasing stability (Fig. 1b). Indeed, for the less stable mutants the interfacial conformational changes are of such a scale that they may even be detected by methods monitoring the overall adsorbed layer thickness. Thus, Fro¨berg et al. investigated the adsorption of wild-type lysozyme and the mutant Ile3 → Trp, where the latter is 2.8 kcal/mol less stable than the former, and found that after adsorption at mica the adsorbed layer thickness for the stability mutant was much ˚ ) than that of the wild type (45–50 A ˚ ) (Fig. 2) [16]. It was also lower (15–17 A observed that the short-range attraction on separation between the adsorbed protein

TABLE 1 Amount of T4 Lysozyme Mutants Adsorbed at Silica and Methylated Silica after (a) Adsorption from 0.01 M Sodium Phosphate Buffer, pH 7.0, for 30 min, (b) Rinsing with Buffer for 30 min after the Initial Adsorption, and (c) Addition of the Cationic Surfactant DTAB for 45 min, Followed by Another Rinsing with Buffer

a

First and second values refer to results obtained for silica and methylated silica, respectively. Source: Ref. 13.



99

FIG. 1 (a) Content of ␣-helix in adsorbed (filled circles) and nonadsorbed (filled diamonds) T4 lysozyme mutants of different structural stability calculated from CD spectra. Shown also is the loss in ␣-helix content on adsorption (open circles). (b) Molecular ellipticity at 222 nm as a functon of time for the wild type (squares), Ile3 → Trp (circles), and Ile3 → Cys (diamonds) adsorbing at silica particles. (Data from Ref. 15.)

layer and pure mica was an order of magnitude larger for the conformationally altered mutant in comparison to the wild type (not shown), indicating a strong effect on the protein–surface interaction. Hence, it seems clear that structural changes do occur in these systems on adsorption to an extent depending on the protein structural stability. Another system which has been investigated in relation to the effect of structural stability on the protein interfacial behavior is human carbonic anhydrase II. In this case, a series of variants of different structural stability can be obtained by © 2003 by Marcel Dekker, Inc.

100

Malmsten et al.

FIG. 2 Force normalized by curvature for wild-type lysozyme (squares) and the Ile3 → Trp T4 lysozyme mutant (circles) adsorbed at mica from water. Only forces measured on approach are shown. (Data from Ref. 16.)

truncation of the N terminus with a certain number of amino acids [19]. Although this system suffers from the molecular weight not being entirely constant for the different stability variants, it has the nice feature of being straightforwardly quantifiable regarding biological activity. It has been found that a decreased activity is reached at lower concentrations of a denaturating cosolute (GuHCl) for the truncated proteins as compared to the intact proteins, which was interpreted as being a consequence of the destabilization of the native structure relative to an intermediate state [19]. Using a series of human carbonic anhydrase II proteins, Billsten et al. investigated the adsorption of this protein at silica using a range of techniques, including CD and fluorescence, and found that as the structural stability of the protein was decreased, the content of ordered structure in the protein after adsorption was reduced, in line with the findings for T4 lysozyme [20]. Furthermore, the active site structure was changed on adsorption for the truncated proteins but not for the intact protein (Fig. 3). Although the interfacial enzymatic activity was not determined, these findings seem to indicate an inverse correlation between enzymatic activity and interfacial conformational changes for this system. This is in line with the frequently observed loss in biological activity on adsorption, e.g., in relation to protein and peptide drugs. Note, however, that the loss in activity and native structure on adsorption by no means is universal. In fact, there are systems which actually depend on these events. Examples of this are lipases, which may have a lower activity in bulk solution than after adsorption at an interface [21,22, and references therein]. Although the existence of a correlation between structural stability in solution, structural loss on adsorption, and interfacial activity seems clear in numerous systems, the origin of these effects is still somewhat unclear. In particular, the relative importance of exposure of hydrophobic groups, affecting the protein–surface, protein–protein, and protein–solvent interactions, entropy gain related to conformational freedom and translational entropy related to hydration water and counterions for both © 2003 by Marcel Dekker, Inc.

FIG. 3 Structural changes in the active site of human carbonic anhydrase II as observed by monitoring the fluorescence of an extrinsic active site probe, DNSA. Shown are the results obtained for the proteins in solution (solid lines) and for the proteins after 24 h equilibration with silica particles (dashed lines) for (a) the pseudo–wild-type protein, (b) Trunc 5 (i.e., four amino acids truncated), and (c) Trunc 17 (i.e., 16 amino acids truncated). (Data from Ref. 20.)


102

Malmsten et al.

the protein and the surface, is unclear at present. As indicated, McGuire et al. have in several studies compared the behavior at silica and methylated silica, and found that both the adsorption and the interfacial conformation depend on the protein structural stability for both these surfaces, but primarily so for the hydrophilic and negatively charged silica surface [13,17]. Furthermore, as already discussed, significant interfacial conformational changes were found by Billsten et al. [15] and Fro¨berg et al. [16] for hydrophilic silica and mica surfaces, respectively. This could be interpreted to indicate that the effects on interfacial conformational change of the protein– surface interaction is of minor importance, since the protein–surface interaction is not likely to be favored by a conformational change and exposure of hydrophobic residues in the case of silica and mica. This, however, is contradicted by the adhesion data from the surface force measurements by Fro¨berg et al. (see above). On the other hand, Su et al. concluded that the limiting adsorption of lysozyme at silica was dependent on the protein–protein interactions [23]. As the protein stability decreases, structural adaptations become more accessible, which may therefore affect the adsorption. This is to some extent analogous to the oligomerization-dependent adsorption extensively discussed in Chapter 12. Regarding the conformational and translational entropy gain processes on adsorption, further systematic studies are in our mind required to reach an improved understanding as to their relative importance.

IV.

INTERFACIAL BEHAVIOR OF INTERACTION MUTANTS

When a homopolymer adsorbs at an interface, it typically loses translational and conformational entropy, since it is not free to move as extensively as in the bulk solution and since the presence of the interface dramatically reduces the number of possible conformations [1]. Consequently, the adsorption of homopolymers is enthalpically driven, and in the absence of a polymer–surface attractive interaction the polymer is depleted from the interface. In the presence of an attraction outbalancing the entropic loss the polymer adsorbs to an extent increasing with the polymer– surface interaction until high adsorption energies, when the adsorption becomes essentially independent of the polymer–surface attraction. Due to the typically large number of monomers in a polymer, even a weak monomer–surface interaction results in a high overall adsorption driving force, and the adsorption is therefore highly cooperative. For proteins, the situation is somewhat more complex, since the adsorption of a protein may result in a conformational entropy gain due to loss in ordered structure. As discussed, it has been concluded that the adsorption of proteins may in some cases be entropically driven, e.g., due to the conformational entropy gain on reduction in the content of ordered conformations or due to gain in entropy related to protein and surface counterions and/or hydrating water molecules. Nevertheless, also for these systems the adsorption could be expected to be favored by an attractive protein–surface interaction. Examples showing this include the higher adsorption for a number of (negatively charged) proteins at hydrophilic and positively charged surfaces than at similarly hydrophilic and negatively charged surfaces, as well as the typically higher adsorption at the latter of similarly charged hydrophilic and hydrophobic surfaces (cf silica and methylated silica) [2,3,24,25]. © 2003 by Marcel Dekker, Inc.


103

In order to illustrate the importance of protein–surface and protein–protein interactions for the adsorption of proteins, Fig. 4 shows the adsorption at methylated silica of proteoheparan sulfate, a glycoprotein consisting of a hydrophobic peptide domain and hydrophilic and highly negatively charged polysaccharide domains [26]. Naturally, the former promotes adsorption at the hydrophobic surface, whereas the adsorption is opposed by electrostatic interactions due to the polysaccharide domains, originating, e.g., from an image charge repulsion between the protein charges and the low-dielectric-constant surface, from a direct electrostatic interaction between the similarly charged protein and surface, and from an electrostatic protein–protein interaction [1,26,27]. Clearly, the adsorption will be determined by the balance in these interactions. From Fig. 4, one can see that at low electrolyte concentration, a limited adsorption occurs at methylated silica, indicating that the repulsive interactions do not totally dominate. For silica, on the other hand, no adsorption was observed. This could be expected, since no hydrophobic adsorption driving force exists in this case. When CaCl2 is added to the system, the adsorption at methylated silica increases notably, which is a consequence of the increased screening of the electrostatic interactions, making the hydrophobic attraction relatively more important. However, also at high electrolyte concentration, no adsorption was observed at methylated silica for the polysaccharide side chains in the absence of the hydrophobic peptide domain [28]. Clearly, in this case the system is effectively below the critical adsorption energy, and the repulsive protein–surface interaction dominates. Although the example indicates the importance of protein–surface and protein– protein interactions for adsorption, the results are not entirely conclusive. In particular, the molecular weight of the polysaccharide side chains is much lower than that of the intact proteoheparan sulfate. Since the adsorption is expected to increase with increasing molecular weight, especially at highly electrostatically screened conditions [1], molecular weight effects cannot be excluded.

FIG. 4 Amount of proteoheparan sulfate adsorbed at methylated silica as a function of time (open circles). At zero time and point (a), proteoheparan sulfate and 1.25 mM CaCl2 were added, respectively. Illustrated also is the (null) adsorption of the heparan sulfate side chains. (Data from Ref. 26.)


104

Malmsten et al.

The structure of proteins in general is governed by a balance of effects, e.g., electrostatic, hydrophobic, and van der Waals interactions; hydrogen bonding; solvation; and entropic effects due to conformational restrictions [2,3,27]. This balance may be quite delicate and thus easily perturbed by changing the solution condition (e.g., pH or electrolyte concentration), temperature, addition of cosolutes (e.g., surfactants, urea, and guanidinium hydrochloride, GuHCl), the presence of an interface (see above), and, notably, mutations. Since pertubations in protein structure and/or structural stability may affect the protein interfacial behavior (discussed above), it is important when investigating the effects of interactions on protein adsorption that the interaction be changed essentially without affecting the protein structure and structural stability. In one of the first attempts to probe these effects, McGuire et al. investigated the effects of net charge and charge location on the adsorption and surfactant-induced elutability of bacteriophage T4 lysozyme at silica and methylated silica with ellipsometry [14]. In this investigation, a series of proteins ranging in net charge from ⫹5 to ⫹9 were obtained by point mutation in either the C- or the N-terminal lobe. Unfortunately, no clear correlation was observed between the protein charge, on one hand, and adsorption and elutability, on the other. Instead, the location of the mutation was found to be of importance. At silica, mutants allowing close proximity of positive charges of the protein and the surface negative charges were found to have a decreased elutability, whereas at methylated silica a decreased elutability was observed for mutations favoring a hydrophobic interaction between the protein and the surface. However, the point mutation approach used did not allow the structural stability to be preserved between the different proteins. Instead, the denaturation free energy spanned the range ⫺1.5 kcal/mol ⱕ ⌬⌬G ⱕ ⫹0.5 kcal/mol compared to the wild type. Since the structural stability has been shown to be of importance for the adsorption of this and other proteins, effects related to this parameter cannot be excluded in the interpretation of the data. In order to investigate the importance of the protein–surface interaction for protein adsorption and interfacial activity, Wannerberger and Arnebrant studied the adsorption, e.g., of the wild type and a mutant of lipase from Humicola lanuginosa, where the latter was modified from the former by an Asp96 → Leu mutation [21,22]. Thus, the mutant is more hydrophobic than the wild type. As can be seen in Table 2, the higher hydrophobicity of the mutant resulted in a higher adsorbed amount at methylated silica compared to the wild type. The effect of the surface hydrophobicity on the adsorption of the two proteins was investigated over a range of contact angles ( nonionic surfactants. Normally, nonionic surfactants do not interact with proteins, which could explain why they seldom have any large cleaning effect on hydrophilic surfaces [35,45]. In addition, on hydrophobic interfaces it has been reported that the head group influences the amount removed. Rapoza and Horbett [4] have observed that the elutability (the relative amount of protein removed) was higher for surfactants with small head groups (SDS, DAHCl, C12E4) than for Tween 20, which has a large sorbitol ring. They also found that the positively charged DAHCl was the most efficient of these four surfactants in removing fibrinogen from polymeric surfaces. The explanation could be that DAHCl has a smaller head group than the other surfactants and thus might adsorb in denser layers to the interface. Similar effects have been seen for BRIJ type of surfactants used as a coating in capillary electrophoresis [46]. It has also been observed that approaching the cloud point of C12E5 increases its competitive power [33]. This might also be related to the packing efficiency of the surfactant. There are indications that at least on hydrophobic surfaces the chain length of the surfactant is of less importance than the head group [4,39,46]. For example, Rapoza and Horbett [4] found no difference in elutability of fibrinogen when investigating alkyl sulfates with varying chain lengths. However, the hydrophobic part of the surfactant might not always be composed of a single straight hydrocarbon chain. Welin-Klintstro¨m [35] has observed that surfactants with complex and bulky hydrophobic parts, such as CHAPS and deoxycholic acid, will be less efficient on hydrophobic and hydrophilic surfaces. Chattoraj and coworkers have studied the influence of surfactant type on the adsorption from mixtures of proteins and surfactants onto interfaces [38,47]. They found that the degree of cooperative adsorption of gelatin and TABs to alumina passed through a maximum with increasing chain length of the surfactant [38]. Similar results have also been observed for fibrinogen and TABs at silicon oxide surfaces [39]. Samanta and Chattoraj studied systems where the presence of surfactant reduced the amount of protein adsorbed [47,48]. For alumina they found that CTAB is more efficient in hindering adsorption of a globular protein (BSA) than SDS [47], while others have found that SDS is better in blocking protein adsorption at silica surfaces [39,49]. The material is not large enough for reaching any general conclusion, but one could speculate that the surfactant is most efficient in hindering protein adsorption when it has the same charge as the surface. Thus, hindering adsorption through solubilization of the protein instead of competition for the interface might be more efficient at hydrophilic interfaces. Similar trends have been observed in the surfactant-induced removal of proteins from solid surfaces [33,35,49]. 2.

Concentrations Below and Above Critical Micelle Concentration

There are few investigations into how the surfactant concentration affects the amount removed [4,32,34,50–52], and from these it is obvious that removal does not occur © 2003 by Marcel Dekker, Inc.

Protein–Surfactant Interactions

327

until a threshold value has been reached. When the threshold has been reached, the removal increases rapidly with increasing surfactant concentration and soon reaches a plateau value. In Fig. 4 the dependence of protein removal on surfactant concentration is illustrated [34]. The threshold value is usually well below the cmc for the surfactant but is probably linked to the cooperative behavior of the surfactant. Factors that would influence the cooperativity, such as surfactant chain length [4] or ionic strength of the solution [4,32], move the onset of desorption in the same direction as they would move the cmc for the surfactant. Furthermore, it is known that the presence of a surface or a polymer usually decreases the critical association concentration with several magnitudes. Samanta and Chattoraj [48], for example, noted that massive binding of SDS to BSA occurred at the alumina interface at a concentration that was nearly 100 times lower than the one expected for cmc. Claesson and coworkers have used surface force measurements to study the interaction between lysozyme adsorbed to mica and anionic surfactants [51,52]. They found that the introduction of the surfactants first leads to an increase in the apparent interfacial charge, and close to cmc they observe a removal of the protein. They did not detect any major conformational changes in the adsorbed proteins due to the adsorption of surfactants [below the critical association constant (cac)]. They also noted that the layer thickness did not increase upon adsorption of surfactants to the protein, indicating that the surfactants have penetrated the protein layer. In the case of removal by displacement Wahlgren and Arnebrant have observed, for lysozyme adsorbed to methylated silica, that the removal followed the adsorption of surfactant to a clean surface [34] (Fig. 4). Note that the start of removal is not only dependent on the surfactant but also on the protein layer. It has been found that the removal started at different concentrations for mutant forms of T4 lysozyme [32] and lipase [50]. The T4 lysozyme investigated had a mutation that gave larger conformational changes of the protein compared to the wild-type protein upon adsorption to interface. This was thought to favor binding of surfactant at low concentrations, leading to solubilization and removal at lower surfactant concentrations than for the wild-type protein [32]. The lipases differed in their hydrophobicities. It was found that removal from methylated

FIG. 4 Dependence of protein removal on surfactant concentration. Removal of lysozyme from methylated silica by C12E5. (Data from Ref. 34.)


328

Wahlgren et al.

silica started at higher surfactant concentrations for the more hydrophobic protein, probability due to stronger interaction between protein and surface [50]. Other investigators have observed that surfactant-induced removal can differ between proteins [4]. It might even differ for the same protein, depending on the adsorption conditions. Horbett and Rapoza have observed a difference in the critical surfactant concentration for removal between fibrinogen layers formed at different concentrations [4]. This was probably due to difference in conformation or orientation of the adsorbed proteins. These issues are discussed in more detail elsewhere in this book. The same concentration dependency is of course also observed for the competitive adsorption of protein and surfactant [37,41,47,48]. The two main differences between competitive adsorption and removal of preadsorbed proteins are that 1.

2.

Proteins and surfactants are present in the solution at the same time, and if the surfactant binds to the protein this might influence the amount of free surfactant molecules, resulting in an apparent shift in cmc and cac of the surfactant. The protein has not adsorbed to the interface before it binds surfactants. This means that protein–surfactant interactions will not be influenced by the presence of the surface.

The adsorption isotherms for proteins and surfactants often differ substantially (Fig. 5). Protein usually adsorbs in a broader concentration interval than surfactants and at lower total concentrations. However, the surfactants are often more surface active than the protein when the concentration is above the critical association concentration. Thus, at least for hydrophobic interfaces the surfactants will start to compete with the protein above this critical concentration, but the proteins will adsorb below it. A practical consequence of this is that if the mixture is diluted the protein might start to adsorb even though the ratio between protein and surfactant is constant [37].

FIG. 5 The competitive adsorption of ␤-lactoglobulin and SDS (squares) and the adsorption of the pure components, SDS (open circles) and ␤-lactoglobulin (crosses), to a methylated silica surface. A degree of dilution of 1 represents a surfactant concentration of 2 cmc in water and a protein concentration of 1 mg/mL. (From Ref. 37.)



B.

Protein

1.

Protein Properties

329

The ability of a surfactant to remove proteins from a surface varies considerably, depending on the protein in question. Due to the complexity of the issue and the large variations between the different proteins, it has been difficult to pinpoint the crucial protein characteristics. Wahlgren et al. [53] made an attempt to study the effect of protein properties on the removal of proteins by dodecyltrimethylammonium bromide (DTAB). The correlations were not very good, but it seemed that at hydrophobic methylated silica surfaces a decrease in the degree or removal might be observed when the size and shell hydrophobicity of the proteins increased. At hydrophilic silicon oxide surfaces some influence of the protein’s size, isoelectric charge, and structural stability was evident. Later studies using mutants of T4 lysozyme have verified the importance of conformational stability both on hydrophilic and hydrophobic silica [54,55]. The effect of charge is, however, far more complicated than what was indicated from earlier studies, and factors such as location of the charge, etc., might play a considerable role [56]. Wannerberger and Arnebrant studied the lipase Humicola lannuginosa and a mutant that had increased hydrophobicity. They found in agreement with the earlier study that increased hydrophobicity decreased the capability of surfactants SDS and C12E5 to remove protein from methylated silica [50]. A more thorough discussion of mutant protein adsorption is given elsewhere in this book. 2.

Structure of the Protein Layer

The structure of the protein layer might also influence the degree of removal by surfactants. It is often observed that only a fraction of the adsorbed proteins can be removed by surfactants. This has been taken as an evidence for a heterogeneous surface layer. The heterogeneity could be due to different orientation of the proteins, different degree of conformational changes of the adsorbed proteins or heterogeneity of the surface [57]. The surface coverage is one of the factors that are thought to influence orientation and conformational changes of the protein molecules. Thus, if the layer formed is below a monolayer there is usually a higher possibility for the protein to adapt to the surface by changing its conformation, and this might lead to lower elutability [4]. However, it has also been observed, in the case of collagen, that high protein density makes the protein more difficult to remove. This could be due to interactions between the adsorbed proteins [58]. A few proteins may form bi- or multilayers when they adsorb to an interface when the bulk concentration is high. This could also influence the removal by surfactants. Figure 6 demonstrates this for the removal of lysozyme by nonionic surfactant C12E5 and SDS. Furthermore, identical experiments have been performed for bovine serum albumin (BSA) (Wahlgren, unpublished data). Lysozyme is known to form bilayers at higher concentrations, while BSA only forms monolayers. When the bulk concentration of the protein is increased, C12E5 seems to become less efficient in removing the lysozyme, but this does not happen to BSA. The decrease seen for lysozyme could thus be due to the formation of a second layer of protein. The formation of bilayers did not seem to affect the cleaning efficiency of SDS. This surfactant binds, in contrast to C12E5, directly to the protein and could remove proteins both by solubization and displacement. © 2003 by Marcel Dekker, Inc.

330

Wahlgren et al.

FIG. 6 Effect of multiple layers of adsorbed protein. Adsorption of lysozyme 1 mg/mL (circles) and 10 mg/mL (squares) followed by desorption by C12E5 (left panel) and SDS (right panel). Adsorption was measured at 25⬚C in a 0.01 M phosphate buffer pH 7. Filled arrows indicate rinsing with buffer; open arrows indicate the addition of surfactant.

There are few studies into the structure of adsorbed protein layers both prior to and after surfactant-induced protein removal. Feng et al. [59] has used atomic force microscopy to study the adsorption and SDS-induced removal of high-density lipoproteins (HDL). They found that HDL adsorbed onto mica surfaces existed both as single proteins and as clusters of two or three particles. After SDS exposure, the number of adsorbed HDL particles had decreased drastically and no clusters were observed. Furthermore the proteins remaining after removal seem to be scattered randomly over the surface. 3. Adsorption Time and Temperature As pointed out, structural changes of the adsorbed protein tend to reduce the amount of protein removed by surfactant. The same is probably true also for restructuring the adsorbed layer. These changes will of course be time [2,5] and temperature dependent [59–62]. Rapoza and Horbett have measured a decrease in surfactantinduced removal of fibrinogen for up to 5 days [5]. The time effects have also been observed on a much faster time scale, i.e., within hours or less from the initial adsorption [3,7]. The time-dependent change in elutability has been correlated to structural changes of the fibrinogen molecule as measured by infrared spectroscopy [7]. It has been observed that the rate of conversion of fibrinogen to a nonelutable state is dependent on the character of the surface [3,7,63,64]. Other proteins, such as lysozyme, BSA, IgG [2], ␤-lactoglobulin [61], and high-density lipoproteins [59], have also shown a time dependence in their elutability. Thus, time-dependent changes are probably a rather general phenomenon for many proteins. McGuire and coworkers used elutability data in combination with simple models for protein adsorption to estimate the conversion of protein from elutable to nonelutable form [54,56,65,66]. They refined the model to contain not only the conversion of the protein to a nonelutable state but also a rate constant for the removal of protein by surfactant [65]. The latter can in some cases be the rate-determining step, and then elutability data will not give any good information on the strength of © 2003 by Marcel Dekker, Inc.


331

protein adhesion to the interface [65]. Thus, elutability data cannot always be used as an indirect measurement of the strength in the interaction between surface and protein. Increased temperature can lead to a higher rate in the time-dependent structural changes of the protein. It can also induce denaturation of the protein and aggregation at the interface and in solution. This will lead to high amounts of protein adsorbed, and the surfactant-induced removal will decrease drastically [59–62] (Fig. 7). The figure illustrates that these effects occur in a narrow temperature interval, as would be expected for temperature-induced protein denaturation. Karlsson et al. [60] have shown that the effect of temperature varies considerably for different surfaces, such as chromium, steel, and methylated silica. For a methylated silica surface the amount removed decreases with increasing temperature between 25 and 80⬚C, although the largest decrease is between 60 and 73⬚C. The metal surface first has an increase in the amount removed by a combination of rinsing and surfactants, when the temperature is increased to 73⬚C. After this the removal decreases when the temperature is further increased. Thus, there might be an optimal temperature for protein removal that differs among different types of surfaces. The temperature has also been shown to effect the competitive adsorption of surfactant and proteins, especially at surfactant concentrations below cmc. Wahlgren et al. [33] observed for fibrinogen and C12E5 that an increase in the temperature from 22 to 34⬚C increased the competitive power of the surfactant. C12E5 has a cloud point around 30⬚C, and it is plausible that the change in the surfactant explains this temperature dependence. C.

Surface

Attempts have been made to correlate surface properties with the degree of surfactant-mediated protein removal. One parameter of interest is, of course, the chemical

FIG. 7 Temperature dependence on the adsorption of 1 mg/mL of ␤-lactoglobulin onto methylated silica followed by removal by twice the cmc of SDS. Filled arrows indicate rinsing with buffer; the open arrow indicates the addition of surfactant.


332

Wahlgren et al.

composition of the material, and Ertel et al. [2] have found for radiofrequency plasma-deposited polystyrene that the change is elutability with time increased with increasing surface oxygen. Other parameters of interest would be surface roughness, surface heterogeneity and the molecular rigidity of, in particular, polymer surfaces. However, it has often been difficult to give simple explanations for the observed results [3,12]. This is not surprising considering that the mechanism of removal, the interaction between surface and surfactant and protein all vary among different surfaces. Still, Horbett and coworkers have successfully correlated the degree of removal induced by SDS (elutability) to the biocompatibility of surfaces [63]. A high degree of fibrinogen removal has been seen for surfaces with good biocompatibility. Some work that best illustrates the complexity of the surface influences has been done by Elwing and coworkers, using surfaces with a hydrophobicity gradient [33,35,64,67]. These gradients are based on silicon oxide surfaces that were made hydrophobic in different degrees by the use of dimethylochlorosilane, rendering one end of the surface strongly hydrophobic (contact angle >90⬚) and one end, where no silanization occurred, strongly hydrophilic (contact angle vitronectin > fibronectin >> albumin. Thus, Steele et al. [3,7,21–24] and others [25,26] have shown that on many (but not all) surfaces, vitronectin adsorption is in excess over fibronectin. Consequently, vitronectin plays a much more important role than fibronectin in mediating adhesion to many surfaces under serum adsorption conditions typical in many cell culture studies. Studies from the author’s lab [27,28] and others [26] with binary mixtures of fibrinogen have shown it typically strongly outcompetes other proteins including albumin and IgG, while fibronectin is only mildly more surface active than these two major plasma proteins, and vitronectin appears to have a surface affinity in between those of fibrinogen and fibronectin. B.

Molecular Potency

Molecular potency is a term coined by J. Steele and used here to mean that the ability of an adhesion protein to mediate adhesive events varies a great deal with how it is adsorbed and is not simply related to the number of adsorbed adhesion molecules per unit area. Thus, for example, fibronectin adsorbed to polystyrene supports cell adhesion, but only at higher adsorbed amounts than when it is adsorbed to the wettable version of polystyrene that has been surface modified for tissue culture [7,29–31]. In addition, coadsorbed albumin enhances the potency of the adsorbed fibronectin [29–32]. Similar variations in the molecular potency of adsorbed fibrinogen have been reported on different surfaces [33], when albumin is coadsorbed [34], or as a function of how long the protein has been on the surface (residence time) [35,36]. These variations in potency are thought to be related to variations in molecular spreading (degree of unfolding) of the adsorbed adhesion proteins, but may also (or instead) be related to variations in the orientation or accessibility of the cell binding domains on the molecules or to how tightly the protein is held to the surface [8,37]. The author has reviewed the relevant literature on molecular potency in greater detail in a prior review [10], and recent examples are discussed in Section V.

IV.

AFFINITY VARIATIONS IN ADSORPTION OF ADHESION PROTEINS: NONFOULING MATERIALS

A.

Types of Nonfouling Materials

The most important recent findings about variations in the affinity of the adhesion proteins have been made by the search for so-called nonfouling materials that would reduce the affinity for most proteins, including the adhesion proteins, to very low levels. Considerable research on materials which would resist protein adsorption and © 2003 by Marcel Dekker, Inc.

Biological Activity of Adsorbed Proteins

397

so be nonfouling has been done in recent years [39–42]. Polysaccharides, lipidlike surfaces including surfaces with phosphorylcholine [43]; self-assembled monolayers (SAMs) on gold substrates made with ethylene glycol–terminated thiol compounds [44–46], hydroxyl-terminated silanes, or addition of polyethylene glycol to the silanated surfaces [47]; and polyethylene glycol are some of the most prominent nonfouling coatings attempted. The resulting coatings sometimes show excellent resistance to protein uptake (‘‘zero’’ protein uptake on some ethylene glycol–terminated self-assembled monolayers on gold [44]), but in other cases still show considerable uptake (e.g., 100 ng/cm2 fibrinogen uptake from plasma to BBA-mPEG coatings on SR in studies by Defife et al. [48] and ca. 160 protein spots detected in SDS washes of PEGylated polycyanoacrylate particles by two-dimensional gel electrophoresis compared to ca. 250 spots in the untreated particles [49]). Lipidlike surfaces also exhibit variable protein repellency. For example, polyurethanes with incorporated phosphorylcholine reduce fibrinogen uptake from buffered fibrinogen solutions to about 30 ng/cm2 [50], while methacryloyloxethyl phosphorylcholine polymers grafted to cellulose hemodialysis membranes reduced total protein adsorption from blood plasma from 1300 to about 300 ng/cm2 [51]. In the following sections, only the poly(ethylene oxide) (PEO)-like materials will be considered further. B.

Poly(Ethylene Oxide)–Like Nonfouling Materials

Protein adsorption to PEO-like materials has been studied extensively in recent years. Many attempts to understand the mechanisms of its low protein adsorption have been made, and several reviews have been written [38–42]. A brief summary of the author’s view of the state of the art is given here. Four topics will be covered: 1. 2. 3. 4.

1.

The mechanisms of nonfouling of PEO The amounts of fibrinogen adsorbed to PEO materials Blood interactions with PEO materials The possible role of PEO in enhancing the biological activity of adhesion proteins adsorbed to PEO surfaces

Mechanisms of PEO Protein Repellency

Steric exclusion and coverage of the surface appear to be the prime factors necessary to prevent protein uptake by PEO-coated surfaces. Steric exclusion arises from the unfavorable entropy that occurs when longer PEO chains become more crowded as protein molecules attempt to fill spaces between adjacent chains. Many early studies supported this idea because longer-chain PEOs often gave lower adsorptivity than lower-molecular-weight versions. However, more recent studies done with tightly packed, self-assembled monolayers have shown that even very short chains terminated in ethylene glycol groups are capable of greatly reducing protein uptake. The SAM studies thus suggest that when the surface coverage is very complete and the layer is consequently very dense, the surface cannot be ‘‘seen’’ by approaching protein molecules. Malmsten and Muller proposed that both steric and shielding effects contribute to protein repellency: ‘‘Once a sufficiently thick adsorbed layer is obtained, attractive interactions between a protein and the underlying surface is essentially fully screened. At this stage, the protein adsorption is typically quite low and only weakly dependent on the PEO molecular weight’’ [52]. © 2003 by Marcel Dekker, Inc.

398

Horbett

Some workers have proposed that PEO mixtures which are polydisperse with respect to molecular weight will have the greatest efficacy in steric stabilization of colloidal particles [53] and in protein repellency [54]. Thus mixtures of long- and short-chain PEO/PPO copolymers adsorbed to gold substrate surface plasmon resonance slides reduced albumin adsorption more than when copolymers of one molecular weight were used [54]. Plasma deposited tetraglyme surfaces probably contain mixtures of PEO chains of varying length, and this may be one of the reasons these materials have shown very good protein repellency. Recent theoretical studies of the mechanisms of protein adsorption suggests which molecular properties of the PEO groups are important to protein repellency [55,56]. In this model, PEO polymers attracted to the surface compete with protein for surface sites and are effective in preventing protein adsorption at equilibrium. In contrast, if the polymer chains are not attracted to the surface, they are expected to be effective in a kinetic sense because the extended chains prevent the close approach of the protein molecules to the substrate. However, at long times the proteins can still sneak through random openings in the extended chains to find a surface site and so these are not good in preventing equilibrium adsorption. This theory also implies the potential utility of surfaces with mixtures of PEO-containing polymeric chains. Morra has provided an overview of theories of fouling resistance in which he stresses the shortcomings of the steric repulsion models [38]. Morra reviews several recent studies showing the likely importance of specific chemical (hydrogen bonding) and conformational (helix versus random coil) features of PEO that the ‘‘physical’’ or steric exclusion theories do not consider. Thus, there is recent evidence that the specific conformation of the PEO chain strongly influences its fibrinogen binding and that some conformers of PEO are actually highly adsorptive to fibrinogen [58]. Implicit in many of the theories of PEO protein repellency is the role of ‘‘structured water’’ created by the strong binding of the water to PEO. Thus, the adsorption of proteins to the PEO chain would require the displacement of a great deal of tightly bound water, effectively causing dehydration of the PEO chains, an event that would be energetically unfavorable as the destruction of all the structured water would create positive entropy. 2.

Fibrinogen Adsorption and Nonfouling Surfaces

Fibrinogen’s relatively high affinity for surfaces and high concentration in blood plasma cause it to adsorb in substantial quantities on most surfaces. Furthermore, we [59] and others [60–62] have observed that even small amounts of adsorbed fibrinogen, much below a monolayer or the levels that normally adsorb, are still quite sufficient to support platelet [59] and monocyte [63] adhesion. Thus, while efforts to prevent cell attachment by using so-called nonfouling surfaces that reduce protein uptake appear to be a fundamentally sound approach to improvement of biocompatibility, it appears it will require the development of extremely good nonfouling technology, i.e., ultralow protein adsorptive surfaces. In reviewing success to date with this approach, it appears that the technology used in many labs does not come very close to this goal. How well do the various types of PEO-based protein resistant materials achieve the design criteria of extremely low fibrinogen uptake? A review of the literature shows that many of the PEO materials have not been evaluated in regard to fibrinogen adsorption, but rather have been evaluated against other proteins. © 2003 by Marcel Dekker, Inc.


399

For example, Irvine et al. used the adsorption of albumin and cytochrome c to demonstrate the protein repellency of their linear and star PEO grafts [64] and many others have similarly used albumin [46,54] or other proteins such as fibronectin [65,66] to characterize repellency. While it is likely that these materials would also exhibit fibrinogen repellency had they been tested, the lack of direct data makes it extremely difficult to assess whether they would achieve the ultralow fibrinogen adsorptivity criteria. Thus, the state of the art is that only a few PEO-like materials have been shown to meet the ultralow fibrinogen uptake criteria, as is now shown. Studies from K. Park’s lab have shown their PEO materials sometimes meet this criteria. Park’s group reported extensive adsorption data from 0.1 mg/mL fibrinogen solutions in buffer to a series of radiation-grafted PEO surfaces [67]. Triblock copolymers (Pluronic娃) of the general formula (ethylene oxide)n –(propylene oxide)m –(ethylene oxide)n with segments of various lengths were used. On most of the PEO-treated glass surfaces, regardless of the PEO chain length, fibrinogen adsorption was below 0.02 ␮g/cm2, compared to 0.47 ␮g/cm2 on the control glass surface. Grafting of Nitinol wire with Pluronic PF127 reduced adsorption to 0.06 ␮g/cm2 compared to 0.5 ␮g/cm2 on Nitinol. Grafting of PF127 to pyrolytic carbon (PC) reduced the adsorption to 0.57 ␮g/cm2 compared to 0.86 ␮g/cm2 on the PC itself. Increasing concentrations of surfactant used during grafting also decreased fibrinogen adsorption, but the lowest adsorption achieved in the concentration studies were approximately 0.05 ␮g/cm2 on glass surfaces treated with 15 mg/mL PP1053. Park’s group later reported fibrinogen adsorption to expanded polytetrafluoroethylene (ePTFE), Silastic, and silanized glass after grafting with a PEO–polybutadiene (PB)–PEO triblock copolymer (COP5000, containing a 5000 MW PEO and 750 MW PB block) that they synthesized [68]. Fibrinogen adsorption to silanized glass grafted with COP5000 was very low (ca. 0.01 ␮g/cm2) but was still substantial on Silastic (ca. 0.07 ␮g/cm2) and ePTFE (ca. 0.25 ␮g/cm2) that had been grafted with COP5000. Malmsten and Muller reported that fibrinogen adsorption to PEO/polylactide copolymers adsorbed to methylated silica was reduced to nearly zero, as detected ellipsometrically, although the protein concentration in the adsorbing solution was not specified [52]. Liu et al. reported bovine plasma fibrinogen adsorption from 4 mg/mL solutions varied from 5.8 to 0.4 ␮g/cm2 (using amide I adsorption obtained from Fourier transform infrared spectroscopy), decreasing as the amount of PEO grafted onto polyurethanes increased [69]. In the latter study, many of the adsorption values are far above a theoretical monolayer and thus very far above the ultralow fibrinogen uptake criteria. Thus the absolute values are somewhat suspect, although the tenfold reduction in fibrinogen uptake is consistent with other studies of PEO surfaces and is probably correct. Kim et al. also reported bovine fibrinogen uptake from 30 mg/mL solutions in buffer onto PEO-grafted polyurethane/polystyrene interpenetrating networks, calculating adsorption from the depletion of the bulk phase by the film samples [70]. Adsorption to the PEO surfaces was reported to be about sevenfold less than the control, but the absolute values reported (ca. 2000 ␮g/cm2) make little sense, as does the use of a depletion method, which is not capable of detecting changes in bulk concentration unless particulate samples are used. Plasma deposited glyme-based materials exhibit ultralow fibrinogen adsorption and platelet adhesion [71]. In studies in our group of fibrinogen adsorption to plasma deposited tetraglyme samples, glow discharge deposition of tetraglyme monomer © 2003 by Marcel Dekker, Inc.

400

Horbett

onto glass, PTFE, and polyethylene was found to cause resistance to fibrinogen uptake and platelet attachment [71]. The resistance was very good, although it did depend on the substrate that was coated. Thus fibrinogen adsorption from 0.1% plasma to tetraglyme-coated PTFE was about 2 ng/cm2, but was about 15 ng/cm2 on tetraglyme-coated glass surfaces. In recent studies in our lab, the earlier work has been verified: tetraglyme-treated fluorinated ethylene-propylene (FEP) showed a 95 to 100% reduction of fibrinogen adsorption, depending on the particular lot of samples used [63]. In these studies, the fibrinogen concentration in the original plasma was quite high (4.39 mg/mL), and the adsorption was done from 1% plasma because the Vroman effect causes maximal adsorption if more diluted plasma is used. Adsorption to FEP controls averaged 116 ng/cm2, which is quite typical for fibrinogen adsorption to hydrophobic polymers. In contrast, fibrinogen adsorption to the five different lots of tetraglyme-treated samples was quite low, varying from 0 to 6.4 ng/cm2. In these studies, the five lots were made under identical conditions, but there was still some variation in fibrinogen adsorption. In preparing coatings that will exhibit ultralow fibrinogen adsorption, we conclude that complete coverage of the substrate with the PEO-like polymerized tetraglyme is important. Second, production of chains of at least three or four ethylene glycol repeat units in a row before being cross-linked or having other nonether carbons is needed for fibrinogen repellency. In addition, retention of a high degree of ether carbon content by minimizing crosslinking is also a critical factor unique to plasma polymerized surfaces. 3. Blood Interactions with PEO Surfaces Although most PEO polymers have exhibited reduced interactions with blood, especially with regard to platelets in in vitro studies, the results have varied widely, and none of these types of materials have yet been shown to achieve high blood compatibility when studied in vivo for longer times. Here a few examples that support these summary statements are given. Detailed reviews of this topic are available [39–42]. Kinam Park’s group reported that PEO-containing triblock copolymers grafted to glass, ePTFE, or Silastic did not exhibit greatly improved blood compatibility in an acute phase canine ex vivo series shunt model when compared to albumin preadsorbed Tygothane control segment, even though in vitro fibrinogen adsorption and in vitro platelet deposition were lowered [68]. Thus they found that platelet deposition to the PEO grafts was about 35% less than the control, whereas much larger reductions in platelet adhesion to these materials had been observed in their in vitro studies. These authors proposed that these differences were likely ascribable to the use of anticoagulation in vitro, as well as temperature and shear effects. Because of the well-known effect of anticoagulants on platelet reactivity, we have avoided their use in our past studies preferring instead to use platelet/red cell suspensions containing physiological calcium and magnesium levels at 37⬚C. Interestingly, Park’s group did report that the platelet deposition in the in vivo study was approximately proportional to the residual fibrinogen adsorption. The latter finding supports the hypothesis that ultralow fibrinogen adsorption will improve blood compatibility. Sung Wan Kim’s group also found that segmented polyurethanes with PEO reduced platelet adhesion in vitro but did not prolong occlusion times when used as arteriovenous shunts in rabbits [72]. In contrast, studies of platelet interactions with © 2003 by Marcel Dekker, Inc.


401

poly(ethylene oxide) networks by Merrill and his coworkers showed that platelet adhesion could be as low as 1 platelet per 1000 ␮m2 after 1 h of blood contact in an ex vivo baboon shunt model, depending on the molecular weight of the PEO and the PEO content of the network [73]. Many other studies have shown reduced platelet adhesion and blood reactivity to PEO surfaces. For example, comblike PEO gradient surfaces made by graft copolymerization of poly(ethylene glycol) monomethacrylate macromers onto corona discharge–treated polyethylene showed decreasing platelet adhesion with increasing graft [74]; PEO attached to polypropylene oxide (PPO) by UV initiated cross-linking reduced platelet adhesion by about 50% compared to PPO [75]; PEO grafted onto Biomer vascular grafts had increased patency compared to Biomer control [76]; and PEO coating of various polymers by a mutual solvent polymeric entanglement method reduced adhesion by up to 20-fold [77]. Wagner’s group used PEO in a novel way by coupling it to fibrinogen already adsorbed to a surface, which resulted in 94–96% reduction of platelet adhesion [78]. In our group, plasma deposited tetraglyme samples exhibiting ultralow fibrinogen uptake have also shown excellent resistance to platelet adhesion in vitro [71]. Some literature suggests that PEO polymers may be thrombogenic in vivo but not thromboadhesive, i.e., the clots they cause are rapidly embolized. In Sefton’s lab, PEO addition to poly(vinyl alcohol) (PVA) grafts reduced protein adsorption, but were ‘‘incapable of reducing the platelet consumptive properties of PVA hydrogel . . .’’ [39,79]. These observations are consistent with older data for a series of acrylic hydrogels varying in water content, where Hanson et al. observed very little adherent clot but increasing rates of platelet consumption on higher-water-content hydrogels [80]. However, neither Sefton’s materials nor the acrylic hydrogels were shown to have ultralow fibrinogen adsorption. The studies do suggest the necessity of studying thromboembolization and platelet consumption in addition to platelet adhesion or clot deposition.

4.

Does PEO Enhance the Biological Activity of Adsorbed Adhesion Proteins?

The biological activity of adsorbed adhesion proteins can be enhanced by the coadsorption of other proteins and also by ‘‘substrate activation’’ (reviewed by the author in Refs. 9 and 10). It is thought that coadsorbed proteins fill in sites on the surface near the adhesion protein, preventing them from completely unfolding as more and more contacts with the surface are made. Substrate activation refers to the enhancement of the biological activity of the adhesion protein when it is adsorbed. Thus, for example, platelets do not bind to fluid phase fibrinogen unless the platelets have first been exposed to an agonist such as ADP or thrombin, while unstimulated platelets bind readily to adsorbed fibrinogen and other adsorbed proteins [81–83]. Two recent studies suggest activation phenomena may occur with PEO and adsorbed adhesion proteins. Thus fibronectin adsorbed to surfaces with moderate amounts of PEO on them cause greater adhesion of fibroblasts than occurs for fibronectin adsorbed to the same surface in the absence of any PEO at all [66]. In addition, cells interacting with fibronectin adsorbed to surfaces with intermediate degrees of PEO exhibited enhanced fibrillar fibronectin deposition by the cells [65]. Liu et al. reported large differences in the percent of denaturation of BSA adsorbed © 2003 by Marcel Dekker, Inc.

402

Horbett

to surfaces carrying various amounts of PEO and that thrombus deposition was minimal at an intermediate degree of PEO content [69]. These observations are potentially of great importance in the design of nonfouling surfaces because they indicate that PEO may actually enhance the ability of adsorbed proteins to interact with cells. Thus PEO surfaces that are incompletely repellant due to incomplete coverage or insufficient steric repulsion may in fact be more reactive with the partner cell than the starting substrate.

V.

VARIATIONS IN MOLECULAR POTENCY OF ADSORBED ADHESION PROTEINS

In this section, studies of the molecular potency of adsorbed adhesion proteins are reviewed, with an emphasis on the more recent literature. Since the main methods to study variations in molecular potency have been the use of antibodies or cells, these topics are reviewed separately. A.

Antibody Binding Studies

Self-assembled monolayers of alkylthiolates on gold have been used in several recent studies to examine the effect of changes in surface chemistry on the biological activity of adsorbed fibronectin. Fibronectin adsorption to methyl or carboxyl terminated monolayers measured with 125I radiolabeled fibronectin was found to vary and was higher on the carboxyl terminated SAMs than on methyl terminated SAMs [84]. The biological activity was studied with a monoclonal antibody to the RGD cell binding domain. As expected from the 125I results, the antibody binding to fibronectin on the methyl terminated SAM was less than that to fibronectin on the carboxyl terminated SAM, but the binding was even lower than expected. Calculation of the fraction of the adsorbed fibronectin able to bind the antibody showed that the ratio of antibody bound to fibronectin adsorbed to the COOH terminated SAM was 25% (mol/mol) compared to 10% on the CH3 terminated SAM surface. When albumin was coadsorbed, the ratios increased to 39 and 21% on the COOH and CH3 surfaces, respectively. These studies, from Grainger’s lab, are summarized in the first entry in Table 1. Another study of antibody binding to fibronectin adsorbed to SAMs terminated in CH3, OH, COOH, or NH2 is summarized in the second entry in Table 1 [57]. In these studies, from Garcia’s lab, it was found that the antibodies employed bound much better to fibronectin adsorbed in low amounts to OH terminated SAM than to the other SAMs tested, although antibody binding to fibronectin on all surfaces reached high levels when high amounts of fibronectin were present on the surfaces. Thus these studies indicate that the epitope to which the antibody binds is present when fibronectin is adsorbed to any of the surfaces, but the antibodies’ binding affinity is much higher for fibronectin adsorbed on the OH SAM than on the other surfaces. The differences in antibody binding affinity were attributed to differences in conformation of the adsorbed fibronectin and were well correlated with differences in cellular responses (reviewed below). As in the first study in Table 1 from Grainger’s lab, fibronectin adsorption was measured with 125I radiolabeled form of the protein. In this study, the binding of one of the antibodies studied (3E3) was also © 2003 by Marcel Dekker, Inc.


403

TABLE 1 Changes in Biological Activity of Adsorbed Adhesion Proteins Detected by Antibody Binding Protein FN FN

FN

Surface CH3 SAM COOH SAM CH3 SAM OH SAM COOH SAM NH2 SAM TCPS, PS

Antibody

Antibody binding behavior

Ref.

III-10 III-10 3E3 HFN7.1

10% of RGD sites bind the Ab 25% of RGD sites bind the Ab Both antibodies ca. 20-fold better to FN on OH SAM than to other SAMs

84

HFN7.1

Bound ca. tenfold better to FN on TCPS than on PS Bound ca. tenfold better to FN on TCPS than on PS Bound about same to FN on both TCPS and PS Bound about same to FN on both TCPS and PS 1.2 (soln.) 2.2 (adsbd.) 2.2 (soln.) 0.6 (adsbd.) 1.8 (soln.) 1.8 (adsbd.) pAb binding to FN is much higher for FN adsorbed in low amounts to TCPS than on PS mAb binding to adsorbed FBGN minimally inhibited by solution FBGN mAb binding to adsorbed FBGN moderately inhibited by solution FBGN mAb binding to adsorbed FBGN not inhibited by solution FBGN

31

pAb 3E1 4B2 FN

Immulon

III-10 III-9 III-4 pAb

FN

TCPS, PS

FBGN

PS

2G-5 (␥ 373-385; RIBS-1)

FBGN

PS

9F9 (␣ 87-100; RIBS-2)

FBGN

PS

DSB2

57

87

97

98

86

85

Note: SAM, self-assembled monoloayers; TCPS, tissue culture polystyrene; PS, polystyrene; FN, fibronectin; FBGN, fibronogen.

greater on the COOH SAM than on the CH3 SAM, as reported in the study from Grainger’s lab, but the differences were much less than observed between the OH and other surfaces. In contrast, the other antibody (HFN7.1) used in the study from Garcia’s lab did not show much difference in binding to fibronectin adsorbed to the COOH and CH3, although it was reported to differ in the study of these surfaces from Grainger’s lab. The third study listed in Table 1 was also done by Garcia and collaborators [31]. It was done with four types of antibodies to fibronectin, three monoclonals, and one polyclonal. Fibronectin was adsorbed to either untreated or tissue culture grade polystyrene. As summarized in the table, one of the monoclonal antibodies and the polyclonal antibody bound much better to fibronectin adsorbed to tissue culture grade polystyrene than to fibronectin adsorbed on plain polystyrene, while © 2003 by Marcel Dekker, Inc.

404

Horbett

the other two monoclonal antibodies bound about the same to fibronectin on either surface. However, as in the studies with SAMs, the differences were evident only at lower amounts of adsorbed fibronectin, i.e., the antibodies bound equally well to fibronectin on all surfaces when fibronectin adsorption was high enough. Thus the differences are not due the complete absence of the epitope for the antibody, but thought by the authors to be due to differences in the affinity of the antibody for the adsorbed fibronectin. The fact that differences in binding to adsorbed fibronectin are noted with some but not all antibodies suggests that the adsorption process affects different regions of the molecule in different ways, e.g., some regions of the adsorbed molecule are more denatured than others. This idea is strongly supported by studies in our laboratory on adsorbed fibrinogen with a panel of ten monoclonal antibodies [36]. In the latter studies, it was found that antibody binding to the adsorbed fibrinogen changed as time elapsed after the protein had been adsorbed, but the changes varied greatly with antibody type. Thus, the binding of some antibodies increased, others decreased, and others remained unchanged with residence time on the surface [36]. Although changes in the affinity of the antibody for adsorbed fibronectin was the explanation given by the Garcia and his collaborators for the fact that antibody binding differences depended on the amount of fibronectin adsorbed to the surface, an alternative explanation is that the differences in antibody binding are due to the presence of two forms of fibronectin on the surface, one that binds the antibody and one that does not. Given the exquisite sensitivity of antibody binding to structural motifs in their antigens, it may be easier to imagine that slight structural alterations completely eliminate the affinity for some of the adsorbed proteins rather than alter the affinity for all of the adsorbed molecules to the same degree. In the all-or-none model, binding of the antibody to the adsorbed fibronectin only occurs at high loading of fibronectin on certain surfaces because most of the molecules loaded at lower densities are indeed unable to bind the antibody, but at higher loadings the adsorbed fibronectin is in a different form due to crowding by previously adsorbed fibronectin molecules. In light of the strong effects of coadsorbed albumin in increasing cell and antibody binding to adsorbed fibronectin, it seems reasonably plausible that at higher fibronectin loadings, some of the adsorbed fibronectin has a similar effect as coadsorbed albumin, preventing the loss of the antibody binding site by preventing complete spreading of the later adsorbed molecules onto the surface. Several other less recent studies with monoclonal antibodies that show the formation of hidden sites after adsorption of fibrinogen [85,86] and fibronectin [87] to surfaces are also summarized in Table 1. In the fourth and fifth studies listed, differences in the amount of binding of antibodies to fibronectin in the surface phase were detected, while in the last three studies it was shown that excess fibrinogen addition to the antibody solution did not prevent the binding of the antibody to adsorbed fibrinogen. In the fourth study of fibronectin listed in Table 1, three different monoclonal antibodies to different domains in fibronectin were used; one bound better to adsorbed fibronectin, one bound less well, and one was the same for adsorbed and dissolved fibronectin. In studies of this type, a preliminary screen of the many different monoclonal antibodies is done to find those that seem to discriminate between surface and dissolved proteins. Only a few of the many antibodies tested show this property, presumably reflecting the fact that only a small subset of the many epitopes possible in proteins is affected by the adsorption process. Clearly, © 2003 by Marcel Dekker, Inc.


405

changes in the adhesion protein occur upon adsorption, but they are not necessarily extensive throughout the entirety of the molecule. Instead, the changes in the adhesion protein are more subtle and limited. B.

Cell Interactions

Variations in the molecular potency of adsorbed adhesion proteins affecting cells have been observed in two distinct experimental settings, and so the author has previously used the terms substrate activation and modulation in connection with these different kinds of studies. The adsorption of adhesion proteins to surfaces often greatly increase their cell adhesive properties, a phenomenon referred to as substrate activation. For example, unstimulated platelets will adhere to adsorbed fibrinogen but platelets do not bind soluble fibrinogen unless the platelets have first been activated with an agonist such as ADP or thrombin [83]. Other types of cells have been shown to bind more avidly to adsorbed fibronectin than to soluble fibronectin (reviewed in Ref. 10). Substrate activation was thought originally to be due to the higher local concentration of adhesion proteins in the adsorbed layer and the consequent increase in binding provided by multiple interactions with the cell. However, recent work suggests that it is due to the exposure of novel binding sites in the adsorbed protein that are normally hidden in the soluble protein, as reviewed above. Related to substrate activation is modulation of the biologic activity of adhesion proteins induced by adsorption on different surfaces. Modulation refers to the fact that chemically different surfaces with similar amounts of an adsorbed adhesion protein exhibit differences in cell attachment or spreading. For example, fibrinogen adsorbed to CF3 rich fluorocarbon gas plasma treated surfaces does not support platelet adhesion as well as fibrinogen adsorbed to tetrafluoroethylene (TFE), even though the surfaces adsorb similar amounts of fibrinogen [33]. Modulation of the degree of interaction of adsorbed fibrinogen with platelets [88,89] and of adsorbed fibronectin with several types of cells [29,32] show that molecular potency of the adhesion protein is affected by surface chemical differences. A more detailed review of modulation effects is available [10]. In the remainder of this section, more recent studies of variations in molecular potency of adhesion proteins adsorbed to surfaces are reviewed in greater detail. Tang’s group recently reported a correlation between phagocyte adhesion after 1-day implantation in the intraperitoneal cavity of mice and the amount of MAC-1 binding sites on fibrinogen adsorbed to five different polymers [90]. MAC-1 is an integrin receptor (also designated CD11b/CD18) on phagocytes that mediates their adhesion to fibrinogen. In these studies, the polymers used were polyethylene terephthalate (PET), polyvinyl chloride (PVC), polyethylene (PE), polydimethylsiloxane (PDMS), and a polyetherurethane (PEU). After 16 h of implantation, large differences in phagocyte accumulation on the surfaces were noted, e.g., the surface bound enzyme activity (used to measure phagocyte accumulation) was 6 on PET, 4 on PVC, 2 on PE, and around 0.5 on PDMS and PEU. In previous studies, Tang and his collaborators had shown that phagocyte accumulation was mediated by adsorbed fibrinogen [19] and that the peptide sequence 190–202 in fibrinogen’s gamma chain, designated P1, is involved in binding of the phagocytes through their MAC1 integrin [91]. Another peptide sequence, 377–395 of the gamma chain (P2), is also involved in binding to MAC-1. Since these epitopes are hidden in soluble fibrinogen, © 2003 by Marcel Dekker, Inc.

406

Horbett

but expressed in fibrin, it was proposed that exposure of these sites makes the implant, with its adsorbed fibrinogen, appear to the phagocytes to be the fibrin normally present at wound site, prompting them to adhere. Using monoclonal antibodies that bind specifically to the P1 or P2 epitopes, Tang and his collaborators studied the state of the adsorbed fibrinogen in several ways. First, they showed that fibrinogen adsorbed to PET bound both the P1 and P2 antibodies, and that the addition of soluble fibrinogen to the antibody solution did not block binding of the antibodies to the adsorbed fibrinogen, although addition of either the P1 peptide or the P2 peptide did block the binding. The inability of soluble fibrinogen to block anti-P1 or anti-P2 binding to the adsorbed fibrinogen confirm that P1 and P2 are neoepitopes that are only exposed after adsorption, another example of substrate activation. The binding of these antibodies to fibrinogen adsorbed to the five polymers also varied a great deal, e.g., anti-P1 binding was high for fibrinogen adsorbed to PET (ca. 0.09) and much lower for fibrinogen adsorbed to PDMS (ca. 0.01). When phagocyte accumulation to the five surfaces was plotted against anti-P1 or anti-P2 binding to fibrinogen adsorbed to these surfaces, there was a roughly linear correlation with reasonably good correlation coefficients (0.9 for P1; 0.70 for P2). These studies provide one of the best and clearest examples of variations in molecular potency of adsorbed adhesion proteins, in the form of both substrate activation (exposure of the P1/P2 sites and phagocyte binding only after fibrinogen is bound) and modulation of the biological activity of the adsorbed protein by differences in surface chemistry (variation in phagocyte and anti-P1/P2 binding to fibrinogen adsorbed to the series of polymers). Garcia and his colleagues have studied the modulation of cell proliferation and differentiation by substrate-dependent changes in fibronectin conformation on both polystyrene based surfaces [31] and on alkylthiolate SAMs [57]. As reviewed above, these authors were able to show differences in the binding of antibodies to fibronectin adsorbed on these surfaces. In the cell studies, they showed myoblast proliferation and differentiation was affected by the state of the adsorbed fibronectin. Although initial cell adhesion and morphology of the myoblast was similar on both plain polystyrene and tissue culture grade polystyrene, after 3 days in culture, the cells on polystyrene had grown to confluence but very few of the cells exhibited the bipolar morphology characteristic of myotube formation. On tissue culture polystyrene, proliferation had reached the subconfluent stage, and more of the cells appeared bipolar. Using an immunofluorescent stain for sarcomeric myosin, a muscle specific marker, only 6% of the cells on polystyrene were musclelike, while 21% of the cells on tissue culture polystyrene were musclelike. Further studies of cellular interactions with these surfaces were done using a novel method that cross-links the integrins of the cell with the adsorbed fibronectin if the integrin and the fibronectin are in fact binding. The cross-linked integrins are collected by SDS extraction and analyzed by gel electrophoresis and antibody staining. The cross-linking studies showed distinct differences in types of cellular integrins engaged by the fibronectin adsorbed to the various surfaces, namely, there was increased engagement of the ␣5␤1 integrin in comparison to the ␣v␤3 integrin by cells interacting with fibronectin on the tissue culture polystyrene in comparison to polystyrene. [The cross-linking studies were done with a fibroblast cell line (IMR-90) rather than myoblasts because integrin expression is constant in fibroblasts, while it varies greatly in the myoblast line as they begin to differentiate. The differentiation-induced changes in integrin expression © 2003 by Marcel Dekker, Inc.


407

would have made the cross-linking studies uninterpretable had myoblasts been used.] The differences in types of cellular integrin engaged by the fibronectin adsorbed on the various surfaces correlated with the changes in differentiation. Since the authors had also shown that function blocking antibodies to ␣5-blocked differentiation of the cells into muscle cells, while an antibody to ␣v did not reduce differentiation, the differences in which type of integrin in the cell was engaged by the fibronectin adsorbed to the two kinds of surfaces and the corresponding differences in degree of differentiation are consistent with a model in which the conformation of the adsorbed fibronectin affects differentiation because it affects the type of integrin engaged. In separate studies, Garcia and Boettiger also showed that the enhanced engagement of the ␣5␤1 integrin by fibronectin results in a proportional increase in the phosphorylation of focal adhesion kinase (FAK), a tyrosine kinase involved in early integrin-mediated signaling [92]. The results thus extend previous studies of modulation effects on cell attachment or spreading to a fairly deep level of differential integrin engagement and consequent changes in cell signaling. In the work from Garcia’s group using SAMs, the integrin cross-linking method was used to show differences in the ␣5 engaged by fibronectin adsorbed by the various SAMs used, and there were also differences in focal adhesion components inside the cells as studied with an immunofluorescent staining method [57]. Finally, this group also showed that cell adhesion strength, as studied with a spinning disk device, was greater when fibronectin was adsorbed to bioactive glasses than on control glasses, despite similar amounts of adsorbed fibronectin, again supportive of the idea that substrate chemistry modulates the molecular potency of the adsorbed fibronectin [93]. Grainger and his colleagues have also done extensive work on the modulation of fibroblast adhesion, spreading, and proliferation on SAMs [84] and on other surfaces [94,95]. As reviewed above, antibody binding to fibronectin adsorbed on COOH and CH3 terminated SAMs showed that the molecular potency of fibronectin was greater when it was adsorbed on the COOH terminated surface. They also found that Swiss 3T3 fibroblast attachment and spreading were greater on the COOH SAM than on the CH3 terminated SAM, provided the fibronectin adsorption was done from serum or in the presence of excess albumin, i.e., surfaces adsorbed with purified fibronectin did not display differences in attachment or spreading. Fibroblast proliferation, done in the presence of serum, was greater on the COOH terminated SAM. Immunofluorescent staining of the cells for filamentous actin, paxillin, and phophotyrosine, markers for focal adhesion formation, was also done. As observed in the attachment and spreading studies, there was a difference in actin and paxillin and phosphorylation only when the surfaces were preadsorbed with serum or mixtures of fibronectin with albumin, and as in the other studies cells on the COOH surface exhibited greater staining for these markers of focal adhesion. The authors concluded that for the COOH SAM, high levels of fibronectin cell binding domain accessibility (detected with an antibody) correlate with high degrees of cell attachment, spreading, and growth, while for the CH3 SAM, lower amounts of fibronectin and cell binding domain availability correlates with lower cell interactions. The authors also found that although coadsorbed albumin enhanced the ability of fibronectin adsorbed to either surface to bind an antibody specific to the cell binding domain of fibronectin, in the case of the CH3 SAM this enhancement was not enough to overcome the accompanying decrease in total amount of adsorbed fibronectin. In other studies of © 2003 by Marcel Dekker, Inc.

408

Horbett

cell interactions with these same SAMs, the same group studied intracellular signaling events by measuring the levels of integrin-regulated GTPase RhoA [96]. Consistent with the effects on cell spreading and growth, the RhoA became approximately twice as activated and membrane localized when the fibroblasts were cultured on the COOH terminated SAM in comparison to the CH3 SAM.

VI.

SUMMARY AND CONCLUSIONS

The biological activity of adsorbed proteins is most clearly expressed by the large and specific effects that adsorption of adhesion proteins have on the interaction of cells with solid surfaces. It is fair to say that the expression of this activity is the major recognition system that allows the body to react to foreign materials that themselves lack any intrinsic recognition motif. Thus the more-or-less accidental and incidental adsorption of the adhesion proteins to the surfaces of implanted biomaterials is nonetheless fundamental to how the body responds to biomaterials. The integrins and their cognate ligands, the adhesion proteins, provide the biochemical underpinnings of cell responses to implanted biomaterials, and form a sound theoretical basis for the development of ultralow protein adsorption materials that would be biocompatible by virtue of completely resisting the uptake of adhesion proteins. However, for most materials, models which attempt to predict or influence the cellular response by considering only the absolute amount of an adhesion protein on the surface are insufficient. Instead, there is abundant evidence to show that the biological activity of the adsorbed adhesion proteins strongly depends on the type of surface to which it adsorbed, so that both the amount and the molecular potency of the adsorbed adhesion protein have to be considered.

REFERENCES 1.

2.

3.

4. 5. 6.

V. Hlady and J. Buijs, Local and global optical spectroscopic probes of absorbed proteins, In Biopolymers at Interfaces (M. Malmsten, ed.). Marcel Dekker, New York, 1998, pp. 181–220. J. L. Brash, Role of plasma protein adsorption in the response of blood to foreign surfaces. In Blood Compatible Materials and Devices (C. P. Sharma and M. Szycher, eds.). Technomic Publishing, Lancaster, 1991, pp. 3–24. J. G. Steele, T. R. Gengenbach, G. Johnson, C. McFarland, B. A. Dalton, P. A. Underwood, R. C. Chateiler, and H. J. Griesser, Mechanism of the initial attachment of human vein endothelial cells onto polystyrene-based culture surfaces and surfaces prepared by radiofrequency plasmas. In Proteins at Interfaces: II. Fundamentals and Applications, Vol. 602 (T. A. Horbett and J. L. Brash, eds.). American Chemical Society, Washington, D.C., 1995, pp. 436–449. C. A. Haynes and W. Norde, Globular proteins at solid/liquid interfaces. Coll. Surf. B: Biointerfaces 2:517–566, 1994. W. Norde, Adsorption of proteins at solid–liquid interfaces. Cells Materials 5:97–112, 1995. M. Mrksich and G. M. Whitesides, Using self-assembled monolayers to understand the interactions of manmade surfaces with proteins and cells. Annu. Rev. Biophys. Biomol. Struct. 25:55–78, 1996.


Biological Activity of Adsorbed Proteins 7.

8. 9. 10. 11.

12. 13.

14.

15. 16. 17. 18.

19. 20.

21.

22.

23.

24.

25.

409

J. G. Steele, B. A. Dalton, G. Johnson, and P. A. Underwood, Polystyrene chemistry affects vitronectin activity: an explanation for cell attachment to tissue culture polystyrene but not to unmodified polystyrene. J. Biomed. Mater. Res. 27:927–940, 1993. T. A. Horbett, The role of adsorbed adhesion proteins in cellular recognition of biomaterials. BMES Bull. 23:5–9, 1999. T. A. Horbett, Principles underlying the role of adsorbed plasma proteins in blood interactions with foreign materials. Cardiovasc. Pathol. 2:137S–148S, 1993. T. A. Horbett, The role of adsorbed proteins in animal cell adhesion. Coll. Surf. B: Biointerfaces 2:225–240, 1994. D. R. Phillips, L. Nannizzi-Alaimo, and K. S. Prasad, ␤3 tyrosine phosphorylation in ␣1b␤3 (platelet membrane GP IIb-IIIa) outside-in integrin signaling. Thromb. Haemostas. 86:246–258, 2001. D. G. Woodside, S. Liu, and M. H. Ginsberg, Integrin activation. Thromb. Haemostas. 86:316–323, 2001. M. D. McKee and A. Nanci, Osteopontin at mineralized tissue interfaces in bone, teeth, and osseointegrated implants: ultrastructural distribution and implications for mineralized tissue formation, turnover, and repair. Microsc. Res. Tech. 33:141–164, 1996. S. Ashkar, G. F. Weber, V. Panoutsakopoulou, M. E. Sanchirico, M. Jansson, S. Zawaideh, S. R. Rittling, D. T. Denhardt, M. J. Glimcher, and H. Cantor, Eta-1 (osteopontin): an early component of type-1 (cell mediated) immunity. Science 287:860–864, 2000. T. A. Horbett, The role of adsorbed adhesion proteins in cellular recognition of biomaterials. BMES Bull. 23:5–9, 1999. W.-B. Tsai and T. A. Horbett, Human plasma fibrinogen adsorption and platelet adhesion to polystyrene. J. Biomed. Mater. Res. 1997. W.-B. Tsai and T. A. Horbett, The role of fibronectin in platelet adhesion to plasma preadsorbed polystyrene. J. Biomater. Sci. Polym. Ed. 10:163–181, 1999. W.-B. Tsai, J. M. Grunkemeier, C. D. McFarland, and T. A. Horbett, Platelet adhesion to polystyrene-based surfaces preadsorbed with plasmas selectivity depleted in fibrinogen, fibronectin, vitronectin or von Willebrand’s Factor. J. Biomed. Mater. Res. 60:348– 359, 2002. L. Tang and J. W. Eaton, Fibrin(ogen) mediates acute inflammatory responses to biomaterials. J. Exp. Med. 178:2147–2156, 1993. J.-P. Xiong, T. Stehle, B. Diefenbach, R. Zhang, R. Dunker, D. L. Scott, A. Joachimiak, S. L. Goodman, and M. A. Arnaout, Crystal structure of the extracellular segment of integrin aVB3. Sciencexpress 1–13, 2001. J. G. Steele, G. Johnson, W. D. Norris, and P. A. Underwood, Adhesion and growth of cultured human endothelial cells on perfluorosulphonate: role of vitronectin and fibronectin in cell attachment. Biomaterials 12:531–539, 1991. J. G. Steele, G. Johnson, and P. A. Underwood, Role of serum vitronectin and fibronectin in adhesion of fibroblasts following seeding onto tissue culture polystyrene. J. Biomed. Mater. Res. 26:861–884, 1992. J. G. Steele, G. Johnson, and P. A. Underwood, Role of serum vitronectin and fibronectin in adhesion of fibroblasts following seeding onto tissue culture polystyrene. J. Biomed. Mater. Res. 26:861–884, 1992. J. G. Steele, G. Johnson, C. Mcfarland, B. A. Dalton, T. R. Gengenbach, R. C. Chatelier, P. A. Underwood, and H. J. Griesser, Roles of serum vitronectin and vitronectin in initial attachment of human vein endothelial cells and dermal fibroblasts on oxygen- and nitrogen-containing surfaces made by radiofrequency plasmas. J. Biomater. Sci. Polym. Ed. 6:511–532, 1994. M. D. Bale, L. A. Wohlfahrt, D. F. Mosher, B. Tomasini, and R. C. Sutton, Identification of vitronectin as a major plasma protein adsorbed on polymer surfaces of different copolymer composition. Blood 74:2698–2706, 1989.


410 26. 27. 28. 29. 30.

31.

32.

33.

34.

35.

36. 37. 38. 39. 40. 41. 42. 43. 44. 45.

46. 47.

Horbett D. J. Fabrizius-Homan and S. L. Cooper, A comparison of the adsorption of three adhesive proteins to biomaterial surfaces. J. Biomater. Sci. Polym. Ed. 3:27–47, 1991. T. A. Horbett, P. K. Weathersby, and A. S. Hoffman, The preferential adsorption of hemoglobin to polyethylene. J. Bioeng. 1:61–78, 1977. T. A. Horbett, Mass action effects on competitive adsorption of fibrinogen from hemoglobin solutions and from plasma. Thromb. Haemostas. 51:174–181, 1984. F. Grinnell and M. K. Feld, Adsorption characteristics of plasma fibronectin in relationship to biological activity. J. Biomed. Mater. Res. 15:363–381, 1981. D. K. Pettit, T. A. Horbett, and A. S. Hoffman, Influence of the substrate binding characteristics of fibronectin on corneal epithelial cell outgrowth. J. Biomed. Mater. Res. 26: 1259–1275, 1992. A. J. Garcia, M. D. Vega, and D. Boettiger, Modulations of cell proliferation and differentiation through substrate-dependent changes in fibronectin conformation. Mol. Biol. Cell 10:785–798, 1999. K. Lewandowska, N. Balachander, C. N. Sukenik, and L. A. Culp, Modulation of fibronectin adhesive functions for fibroblasts and neural cells by chemically derivatized substrata. J. Cell. Physiol. 141:334–345, 1989. D. Kiaei, A. S. Hoffman, T. A. Horbett, and K. R. Lew, Platelet and monoclonal antibody binding to fibrinogen adsorbed on glow-discharge-deposited polymers. J. Biomed. Mater. Res. 29:729–739, 1995. J. A. Chinn, S. E. Posso, T. A. Horbett, and B. D. Ratner, Postadsorptive transitions in fibrinogen adsorbed to polyurethanes: changes in antibody binding and sodium dodecyl sulfate elutability. J. Biomed. Mater. Res. 26:757–778, 1992. J. A. Chinn, S. E. Posso, T. A. Horbett, and B. D. Ratner, Post-adsorptive transitions in fibrinogen adsorbed to biomer: changes in baboon platelet adhesion, antibody binding, and sodium dodecyl sulfate elutability. J. Biomed. Mater. Res. 25:535–555, 1991. T. A Horbett and K. R. Lew, Residence time effects on monoclonal antibody binding to adsorbed fibrinogen. J. Biomater. Sci. Polym. Ed. 6:15–33, 1994. J. A. Chinn, R. E. Phillips, K. R. Lew, and T. A. Horbett, Tenacious binding of fibrinogen and albumin to pyrolite carbon. J. Coll. Interf. Sci. 184:11–19, 1996. M. Morra, On the molecular basis of fouling resistance. J. Biomater. Sci. Polym. Edn. 11:547–570, 2000. G. R. Llanos and M. V. Sefton, Review: does polyethylene oxide possess a low thrombogenicity? J. Biomater. Sci. Polym. Ed. 4:381–400, 1993. D. Leckband, S. Sheth, and A. Halperin, Grafted poly(ethylene oxide) brushes as nonfouling surface coatings. J. Biomater. Sci. Polym. Ed. 10:1125–1147, 1999. J. H. Lee, H. B. Lee, and J. D. Andrade, Blood compatibility of polyethylene oxide surfaces. Prog. Polym. Sci. 20:1043–1079, 1995. A. S. Hoffman, Non-fouling surface technologies. J. Biomater. Sci. Polym. Ed. 10:1011– 1014, 1999. K. Ishihara and Y. Iwasaki, Reduced protein adsorption on novel phospholipid polymers. J. Biomater. Applications 13:111–127, 1998. K. L. Prime and G. M. Whitesides, Self-assembled organic monolayers: model systems for studying adsorption of proteins at surfaces. Science 252:1164–1167, 1991. R. L. C. Wang, H. J. Kreuzer, and M. Grunze, Molecular conformation and solvation of oligo(ethylene glycol)-terminated self assembled monolayers and their resistance to protein adsorption. J. Phys. Chem. B 101:9767–9773, 1997. H. B. Lu, C. T. Campbell, and D. G. Castner, Attachment of functionalized poly(ethylene glycol) films to gold surfaces. Langmuir 16:1711–1718, 2000. G. Mao, D. G. Castner, and D. W. Grainger, Polymer immobilization to alkylchlorosilane organic monolayer films using sequential derivatization reactions. Chem. Mater. 9:1741– 1750, 1997.



49.

50.

51.

52. 53. 54.

55.

56. 57.

58.

59. 60. 61.

62. 63.

64.

65.

411

K. M. DeFife, M. S. Shive, K. M. Hagen, D. L. Clapper, and J. M. Anderson, Effects of photochemically immobilized polymer coatings on protein adsorption, cell adhesion and the foreign body reaction to silicone rubber. J. Biomed. Mater. Res. 44:298–307, 1999. M. T. Paracchia, S. Harnisch, H. Pinto-Alphandary, A. Gulik, J. C. Dedieu, D. Desmaele, J. d’Angelo, R. H. Muller, and P. Couvreur, Visualization of in vitro protein-rejecting properties of PEGylated stealth polycyanoacrylate nanoparticles. Biomaterials 20:1269– 1275, 1999. L. Ruiz, E. Fine, J. Voros, S. A. Makohliso, D. Leonard, D. S. Johnston, M. Textor, and H. J. Mathieu, Phosphorylchlorine-containing polyurethanes for the control of protein adsorption. J. Biomater. Sci. Polym. Ed. 10:931–955, 1999. K. Ishihara, T. Shinozuka, Y. Hanazaki, Y. Iwasaki, and N. Nakabayashi, Improvement of blood compatibility on cellulose hemodialysis membrane: IV. Phospholipid polymer bonded to the membrane surface. J. Biomater. Sci. Polym. Ed. 10:271–282, 1999. M. Malmsten and D. Muller, Interfacial behavior of ‘new’ poly(ethylene oxide)-containing copolymers. J. Biomater. Sci. Polym. Ed. 10:1075–1087, 1999. V. Stenkamp and J. Berg, The role of long tails in steric stabilization and hydrodynamic layer thickness. Langmuir 13:3827–3832, 1997. K. D. Pavey and C. J. Olliff, SPR analysis of the total reduction of protein adsorption to surfaces coated with mixtures of long- and short-chain polyethylene oxide block copolymers. Biomaterials 20:885–890, 1999. T. B. McPherson, A. Kidane, I. Szleifer, and K. Park, Prevention of protein adsorption by tethered poly(ethylene oxide) layers: experiments and single-chain mean-field analysis. Langmuir 14:176–186, 1998. J. Satulovsky, M. A. Carignano, and I. Szleifer, Kinetic and thermodynamic control of protein adsorption. Proc. Natl. Acad. Sci. USA 97:9037–9041, 2000. B. G. Keselowsky, D. M. Collard, and A. J. Garcia, Surface chemistry modulates fibronectin conformation and directs integrin binding and specificity to control cell adhesion, submitted (2002). P. Harder, M. Grunze, R. Dahint, G. M. Whitesides, and P. E. Laibinis, Molecular conformation in oligo(ethylene glycol)-terminated self-assembled monolayers on gold and silver surfaces determines their ability to resist protein adsorption. J. Phys. Chem. B 102:426–436, 1998. W.-B. Tsai, J. M. Grunkemeier, and T. A. Horbett, Human plasma fibrinogen adsorption and platelet adhesion to polystyrene. J. Biomed. Mater. Res. 44:130–139, 1999. K. Park, F. W. Mao, and H. Park, Morphological characterization of surface-induced platelet activation. Biomaterials 11:24–31, 1990. K. Park, F. W. Mao, and H. Park, The minimum surface fibrinogen concentration necessary for platelet activation on dimethyldichlorosilane-coated glass. J. Biomed. Mater. Res. 25:407–420, 1991. G. A. Adams and I. A. Feuerstein, How much fibrinogen or fibronectin is enough for platelet adhesion? Trans. Am. Soc. Artif. Int. Organs 27:219–224, 1981. M. Shen, Y. V. Pan, M. S. Wagner, K. D. Hauch, D. G. Castner, B. D. Ratner, and T. A. Horbett, Inhibition of monocyte adhesion and fibrinogen adsorption on glow discharge plasma deposited tetraethylene glycol dimethyl ether. J. Biomater. Sci. Polym. Ed. 12:961–978, 2001. D. J. Irvine, A. M. Mayes, S. K. Satija, J. G. Barker, S. J. Sofia-Allgor, and L. G. Griffith, Comparison of tethered star and linear poly(ethylene oxide) for control of biomaterials surface properties. J. Biomed. Mater. Res. 40:498–509, 1998. G. Altankov, V. Thom, T. Groth, K. Jankova, G. Jonsson, and M. Ulbright, Modulating the biocompatibility of polymer surfaces with poly(ethylene glycol): effect of fibronectin. J. Biomed. Mater. Res. 52:219–230, 2000.


412 66.

67.

68.

69.

70.

71.

72. 73.

74. 75.

76.

77.

78.

79.

80.

81.

82. 83.

Horbett V. Thom, G. Altankov, T. Groth, K. Jankova, G. Jonsson, and M. Ulbricht, Optimizing cell–surface interactions by photografting of poly(ethylene glycol). Langmuir 16:2756– 2765, 2000. T. B. McPherson, H. S. Shim, and K. Park, Grafting of PEO to glass, Nitinol, and pyrolytic carbon surfaces by gamma irradiation. J. Biomed. Mater. Res. 38:289–302, 1997. A. Kidane, G. C. Lantz, S. Jo, and K. Park, Surface modification with PEO-containing triblock copolymer for improved biocompatibility: in vitro and ex vivo studies. J. Biomater. Sci. Polym. Ed. 10:1089–1105, 1999. S. Q. Liu, Y. Ito, and Y. Imanishi, Synthesis and non-thrombogenicity of polyurethanes with poly(oxyethylene) side chains in soft segment regions. J. Biomater. Sci. Polym. Ed. 1:111–122, 1989. J. H. Kim, M. Song, H. W. Roh, Y. C. Shin, and S. C. Kim, The in vitro blood compatibility of poly(ethylene oxide)-grafted polyurethane/polystyrene interpenetrating polymer networks. J. Biomater. Sci. Polym. Ed. 11:197–216, 2000. G. P. Lopez, B. D. Ratner, C. D. Tidwell, C. L. Haycox, R. J. Rapoza, and T. A. Horbett, Glow discharge plasma deposition of tetraethylene glycol dimethyl ether for foulingresistant biomaterial surfaces. J. Biomed. Mater. Res. 26:415–439, 1992. K. D. Park, W. G. Kim, H. Jacobs, T. Okano, and S. W. Kim, Blood compatibility of SPUU-PEO-heparin graft copolymers. J. Biomed. Mater. Res. 26:739–756, 1992. E. L. Chaikof, E. W. Merrill, J. E. Coleman, K. Ramberg, R. J. Connolly, and A. D. Callow, Platelet interaction with poly(ethylene oxide) networks. AIChE J. 36:994–1002, 1990. J. H. Lee, B. J. Jeong, and H. B. Lee, Plasma protein adsorption and platelet adhesion onto comb-like PEO gradient surfaces. J. Biomed. Mater. Res. 34:105–114, 1997. J. G. F. Bots, A. Bantjes, and L. van der Does, Small diameter blood vessel prostheses from blends of polyethylene oxide and polypropylene oxide. Biomaterials 7:393–399, 1986. C. Nojiri, T. Okano, H. A. Jacobs, K. D. Park, S. F. Mohammed, D. B. Olsen, and S. W. Kim, Blood compatibility of PEO grafted polyurethane and HEMA/styrene block copolymer surfaces. J. Biomed. Mater. Res. 24:1151–1171, 1990. N. P. Desai and J. A. Hubbell, Solution technique to incorporate polyethylene oxide and other water-soluble polymers into surfaces of polymeric biomaterials. Biomaterials 12: 144–153, 1991. C. R. Deible, P. Petrosko, P. C. Johnson, E. J. Beckman, A. J. Russell, and W. R. Wagner, Molecular barriers to biomaterial thrombosis by modification of surface proteins with polyethylene glycol. Biomaterials 20:101–109, 1999. G. R. Llanos and M. V. Sefton, Immobilization of poly(ethylene glycol) onto a poly(vinyl alcohol) hydrogel: 2. Evaluation of thrombogenicity. J. Biomed. Mater. Res. 27:1383–1391, 1993. S. R. Hanson, L. A. Harker, B. D. Ratner, and A. S. Hoffman, In vivo evaluation of artificial surfaces with a nonhuman primate model of arterial thrombosis. J. Lab. Clin. Med. 95:289–304, 1980. B. S. Coller, Interaction of normal, thrombasthenic, and bernard-soulier platelets with immobilized fibrinogen: defective platelet-fibrinogen interaction in thrombasthenia. Blood 55:169–178, 1980. J. A. Chinn, T. A. Horbett, and B. D. Ratner, Baboon fibrinogen adsorption and platelet adhesion to polymeric materials. Thromb. Haemostas. 65:608–617, 1991. B. Savage and Z. M. Ruggeri, ‘‘Selective recognition of adhesive sites in surface-bound fibrinogen by glycoprotein IIb-IIIa on nonactivated platelets. J. Biol. Chem. 266:11227– 11233, 1991.



85.

86.

87.

88.

89. 90. 91. 92. 93.

94.

95.

96. 97.

98.

413

K. B. McClary, T. P. Ugarova, and D. W. Grainger, Modulating fibroblast adhesion, spreading, and proliferation using self-assembled monolayer films of alkylthiolates on gold. J. Biomed. Mater. Res. 50:428–439, 2000. J. Soria, C. Soria, M. Mirshahi, C. Boucheix, A. Aurengo, J. Y. Perrot, A. Bernadou, M. Samama, and C. Rosenfeld, Conformational change in fibrinogen induced by adsorption to a surface. J. Coll. Interf. Sci. 107:204–208, 1985. T. P. Ugarova, A. Z. Budzynski, S. J. Shattil, Z. M. Ruggeri, M. H. Ginsberg, and E. F. Plow, Conformational changes in fibrinogen elicited by its interaction with platelet membrane glycoprotein GPIIb-IIa. J. Biol. Chem. 268:21080–21087, 1993. T. P. Ugarova, C. Zamarron, Y. Veklich, R. D. Bowditch, M. H. Ginsberg, J. W. Weisel, and E. F. Plow, Conformational transitions in the cell binding domain of fibronectin. Biochemistry 34:4457–4466, 1995. J. N. Lindon, G. McManama, L. Kushner, E. W. Merrill, and E. W. Salzman, Does the conformation of adsorbed fibrinogen dictate platelet interactions with artificial surfaces? Blood 68:355–362, 1986. J. H. Silver, H.-B. B. Lin, and S. L. Cooper, Effect of protein adsorption on the bloodcontacting response of sulphonated polyurethanes. Biomaterials 14:834–844, 1993. W. J. Hu, J. W. Eaton, and L. Tang, Molecular basis of biomaterial-mediated foreign body reactions. Blood 98:1231–1238, 2001. L. Tang, T. P. Ugarova, E. F. Plow, and J. W. Eaton, Molecular determinants of acute inflammatory responses to biomaterials. J. Clin. Invest. 97:1329–1334, 1996. A. J. Garcia and D. Boettiger, Integrin–fibronectin interactions at the cell–material interface: initial integrin binding and signaling. Biomaterials 20:2427–2433, 1999. E. Heilmann, P. Friese, S. Anderson, J. N. George, S. R. Hanson, S. A. Burstein, and G. L. Dale, Biotinylated platelets: a new approach to the measurement of platelet life span. Br. J. Haematol. 85:729–735, 1993. D. W. Grainger, G. Pavon-Djavid, V. Migonney, and M. Josefowicz, Assessment of fibronectin conformation adsorbed to polytetrafluoroethylene surfaces from serum protein mixtures and correlation to support of cell attachment in culture. J. Biomater. Sci. Polym. Ed. (in press), 2001. A. L. Koenig, V. Gambillara, and D. W. Grainger, Correlating fibronectin adsorption with endothelial cells adhesion and signaling on polymer substrates. J. Biomed. Mater. Res. (in press), 2001. K. B. McClary and D. W. Grainger, RhoA-induced changes in fibroblasts cultured on organic monolayers. Biomaterials 20:2435–2446, 1999. F. Grinnell and M. K. Feld, Fibronectin adsorption on hydrophobic and hydrophilic surfaces detected by antibody binding and analyzed during cell adhesion in serum containing medium. J. Biol. Chem. 257:4888–4893, 1982. C. Zamarron, M. H. Ginsberg, and E. F. Plow, Monoclonal antibodies specific for a conformationally altered state of fibrinogen. Thromb. Haemostas. 64:41–46, 1990.


16 Protein Interactions with Monolayers at the Air–Water Interface DAVID W. BRITT

Utah State University, Logan, Utah, U.S.A.

G. JOGIKALMATH and VLADIMIR HLADY Salt Lake City, Utah, U.S.A.

I. A.

University of Utah,

INTRODUCTION: LANGMUIR FILMS AS MEMBRANE MIMICS Motivation

Proteins vary widely in size, shape, and function, yet they all have the common feature of being surface-active macromolecules. This high interfacial activity arises from the amphipathic nature of proteins, which are composed of polar, hydrophobic, and charged residues. When left unchecked, they accumulate at interfaces, reaching surface concentrations far greater than the corresponding bulk concentration. In most applications this is an undesirable property leading to fouling of surfaces and possible denaturation, aggregation, and loss of activity of the proteins. Inside living organisms, protein surface activity is primarily controlled by the properties of the surfaces (mainly lipid membranes) which they encounter. Understanding how protein adsorption is regulated on biological membranes and attempting to mimic these systems in the laboratory are topics relevant to all medical and biotechnology fields and are major focuses of current research [1–3]. This chapter reviews protein adsorption to Langmuir monolayers, emphasizing the advantages of using insoluble amphiphile films as model membranes and illustrating the benefits offered by mixed monolayers. Methods of tailoring film properties to promote or discourage protein adsorption, as depicted in Fig. 1, are introduced. The first half of the chapter reviews Langmuir monolayers, protein–monolayer interactions, and relevant instrumentation. This is followed by the presentation of a novel means of enhancing the electrostatic binding of proteins to mixed charged/ neutral amphiphile monolayers. The remainder of the chapter is dedicated to the socalled protein-repellent surfactant films, focusing on the influence of monolayer packing density, surfactant conformation, and osmotic (steric) barriers to protein adsorption.

B.

Model Membranes

The term Langmuir monolayer describes a floating surfactant film situated at an air– water or oil–water interface. These films were first characterized in the late 1800s © 2003 by Marcel Dekker, Inc.

416

Britt et al.

FIG. 1 Tailoring monolayer properties to control protein adsorption. Optimizing the electrostatic binding of a charged protein to an oppositely charged monolayer by diluting the charged amphiphile in a neutral amphiphile matrix. The key is to select the right ratio of charged and neutral amphiphiles (upper panels). A PS-PEO monolayer can be changed from a proteinadsorbing to protein-repellent form by increasing the surface density of PS-PEO and thus forcing PEO chains into a brush configuration (lower panels).

in the kitchen of Agnes Pockles, who studied soap films using a water-filled pan with a compression barrier and primitive surface balance—the predecessor of the modern Langmuir trough. This technology rapidly came to maturity in the early to mid-1990s through work by Katherin Blodgett and Irving Langmuir, who developed methods of transferring films to solid supports and performed the preliminary investigations of protein films at the air–water interface [4]. In addition to Langmuir monolayers, a variety of membrane mimics, such as black lipid membranes, supported bilayers, and single- and multilamellar vesicles, have been employed in protein–membrane investigations. While vesicles are perhaps the most popular systems, they have some fundamental limitations with regard to controlling amphiphile packing density, membrane composition, phase behavior, and curvature [5]. The planar geometry of black lipid membranes [6,7] and supported bilayers [8] eliminates the curvature problem, and these films are superior for integral © 2003 by Marcel Dekker, Inc.

Protein–Monolayer Interactions

417

membrane–protein studies and measuring membrane permeability, yet they do not offer the simplicity, stability, or control over film properties attainable with Langmuir monolayers. The main deficit of Langmuir films as membrane mimics is their asymmetry, representing only one-half the bilayer leaflet. But with this limitation comes certain advantages. For instance, many amphiphiles (phospholipids, fatty acids, polymers, proteins, etc.) that cannot be prepared as vesicles can be spread as Langmuir films. Moreover, in contrast to bilayers, monolayer membranes lend themselves to the direct potentiometric measurement of the surface potential [5,9,10], while the Langmuir trough barrier system allows the packing density and phase state of the films to be precisely controlled. The continued popularity of Langmuir monolayers as membrane mimics is a testament to their versatility and unique properties [5,11,12].

II.

LANGMUIR FILMS AND INSTRUMENTATION

A.

Monolayer Characterization: Surface Pressure and Surface Potential

The collective response of insoluble amphiphiles at the air–water interface as they are compressed from a disordered, two-dimensional ‘‘gaslike’’ state through any phase transitions to a close-packed condensed state provides a unique signature in the form of the well-known surface pressure (␲) versus molecular area (A) isotherm. The ␲-A isotherm, usually measured with a surface balance (Wilhelmy plate), provides an indication of interamphiphile as well as amphiphile subphase interactions. For mixed amphiphile films miscibility diagrams can be constructed from a series of isotherms collected for various ratios of the two amphiphiles [13]. Such diagrams highlight any deviation of mixed film properties from ‘‘ideal’’ additive mixing behavior. For molecular area, deviation from ideality upon mixing (⌬Amix) is calculated at a given surface pressure according to ⌬Amix(␲) = A12(␲) ⫺ [␹1A1(␲) ⫹ ␹2A2(␲)]

(1)

where A12 is the average amphiphile molecular area measured for the mixture at a given surface pressure, and A1 and A2 are the molecular areas of the pure components at the same surface pressure. Multiplying A1 and A2 by their respective mole fractions in the mixture, ␹1 and ␹2, yields their expected contributions to the net area based on ideal mixing. Additional parameters such as surface potential (⌬V) and effective dipole moment (␮) can also be evaluated in the same manner to yield ⌬⌬Vmix and ⌬␮mix. The surface potential provides an indication of amphiphile dipole orientation (alkyl-tail tilt angle), lateral density, and charge state. Moreover, ⌬V contains contributions from oriented water dipoles, ions in the electric double layer or proteins accumulated beneath the film. The vibrating plate condenser (Kelvin method) is the most common means to measure ⌬V. In this method, a capacitance developed across the interface separating a platinum electrode in the subphase and an oscillating electrode held 1–2 mm above the water surface is compensated by an applied DC voltage [9,10]. ⌬V can be normalized in terms of molecular area to remove the influence of amphiphile packing density and expressed as the effective dipole moment according to the Helmholtz equation: © 2003 by Marcel Dekker, Inc.

418

Britt et al.

␮ = ␧o⌬VA

(2)

where ␧O is the permittivity of the vacuum, and A is the average area available per molecule (determined from the ␲-A isotherm). It must be stressed that ␮ is not an absolute value, rather it is an effective dipole moment, resulting from a change of the mean dipole moment density normal to the interface related to one amphiphile with respect to the clean water surface. Spreading a monolayer on a clean air–water interface creates two new interfaces, namely, the alkyl-tail–air interface and the head group–water interface. Due to the low dielectric surrounding the alkyl-tail terminal methyl groups, their contribution to the effective dipole moment is quite large: for close-packed hydrocarbon tails, Vogel and Mo¨bius calculated this contribution to be 0.35 D (corresponding to 660 mV of the measured surface potential) [9]. Consequently, any changes in the alkyltail tilt angle can lead to significant changes in ⌬V and ␮, as discussed in Section III.D. When considering the interactions of proteins or other molecules with amphiphile head groups, the contribution of the alkyl-tail–air interface (or for transferred films, the alkyl-tail–substrate interface) is often ignored. However, as demonstrated by Cordroch et al., the chemistry and local environment of the terminal groups of the alkyl-tails can dramatically shift the protonation equilibria of head groups by 1– 2 units [14,15], which could certainly influence protein interactions at the head group region. B.

Protein Interfacial Activity and Surface Tension

Before discussing protein–monolayer interactions, we first consider protein adsorption to a monolayer-free interface. In this case, a decrease in the initial surface tension from that of the pure liquid, ␥o (⬃72 mN/m for water), to that of liquid plus protein, ␥p, provides an indication of protein surface activity. This difference in surface tension is the surface pressure, ␲:

␲ = ␥o ⫺ ␥p

(3)

Surface pressure arises from the lowered chemical potential of the surface layer of water as it mixes with proteins (or other surfactants). Bulk water molecules diffuse to the region of lower chemical potential, bringing about the increase in surface pressure. If a monolayer of amphiphiles is already present at the interface, then the protein must work against the surface pressure exerted by this film. In this case the net surface pressure is given by

␲ = ␥o ⫺ ␥m ⫺ ␥p

(4)

where ␥m is the surface tension of the interface plus monolayer. Note that the term ␥o ⫺ ␥m is the initial surface pressure (␲o) due to the monolayer, which determines the extent (␥p) to which protein can penetrate the film. For close-packed films (high ␲o), the surface balance has the major disadvantage of being insensitive to proteins that adsorb, but cannot penetrate or otherwise induce a change in the film. It has been calculated that a protein needs some 10 to ˚ 2 (6–10 amino acid residues) cleared at the air–water interface in order to 15 A penetrate the film [16]. For this reason measurements are often restricted to less © 2003 by Marcel Dekker, Inc.


419

densely packed films (low ␥m), which permit some extent of protein insertion. When working with close-packed films, techniques such as surface plasmon resonance or light reflection must be used, as discussed next. C.

Light Reflection at the Air–Water Interface

Detecting proteins that adsorb to a Langmuir monolayer but do not induce a change in surface pressure requires more advanced methods (usually optical or gravimetric) than the surface balance. Here we discuss two optical techniques used in our laboratory, light reflection and surface plasmon resonance, for monitoring protein interfacial activity. Light reflection has traditionally been used for measuring the spectra of dye molecules at the air–water interface as a function of packing density and subphase conditions [17]. Fixed wavelength adsorption kinetics of soluble light-absorbing molecules, such as proteins, to the interface can also be attained with this technique [18]. The enhancement of reflection at normal incidence from the interface due to the presence of dye (or protein) is given by [17] ⌬R = A(Ri)0.5 ⫹ ␳2

(5)

where Ri is the reflectivity of the monolayer free interface, ␳ the reflectivity of the monolayer plus protein interface, and A the absorption of the protein. It is often the case that ␳2 is negligible compared to A(Ri)0.5, in which event reflection is equivalent to absorption. Since the contribution of protein in the subphase is accounted for in the reference section of the trough, sensitivity is restricted to protein adsorbed at the interface, as depicted in Fig. 2. For such a differential measurement, it is important to have protein evenly distributed in the subphase. Injecting the protein into a stirred subphase does not 2

FIG. 2 Measuring protein–monolayer interactions as a difference in light reflection between the monolayer-containing interface and a monolayer-free reference.


420

Britt et al.

lead to a rapid and uniform distribution of protein under both the reference and measuring sections of the trough. Thus, for the reflectivity data presented in this chapter, a 0.01 mg/mL protein (horse spleen ferritin) solution was gently poured into a 18 ⫻ 56 cm2 Teflon trough, the surface aspirated clean, and a baseline measurement made at 360 nm. Next, the monolayer was spread from chloroform and rapidly compressed to the desired packing density while measuring ⌬R [18]. D.

Surface Plasmon Resonance at the Air–Water Interface

Since another chapter in this monograph is devoted to the surface plasmon resonance (SPR) technique (Chapter 22), we limit discussion to our adaption of this method to monitoring protein interaction with insoluble poly(styrene)–poly(ethylene oxide) (PS-PEO) block copolymer Langmuir films. Commercial Spreeta SPR sensors (Texas Instruments, Inc.) were used for SPR measurements. These devices combine the sensor surface with all the optic and electronic components required for SPR experiments in a compact assembly with polarized near-infrared light (840 nm) as the light source. The gold surface of the SPR sensor was first cleaned in an oxygen plasma and then modified by octadecanethiol (2 mM in absolute ethanol for 10 min), followed by rinsing with absolute ethanol and then double-distilled water. Prior to the protein adsorption experiment, the hydrophobic SPR sensor was preinitialized in air, calibrated in double-distilled water, and an initial reading of refractive index (RIU) obtained by measuring the PBS buffer solution used as the subphase. The PS-PEO monolayer was then prepared on this subphase and the sensor lowered into full contact with the monolayer. The SPR signal was first recorded as a baseline for at least 10 min to ensure the integrity of the monolayer as well as the stability of the signal (in control experiments a stable signal from the PS-PEO monolayer was obtained over 5 h). Once the baseline signal was recorded, 5 mL of protein (HSA, 1 mg/mL) was injected into the PBS subphase where a small Teflon stir bar mixed the subphase and helped HSA toward the surface, as depicted in Fig. 3. The final HSA concentration in the PBS subphase was 0.008 mg/mL. The SPR signal acquisition was initiated simultaneously with the protein injection and, in parallel, a change of surface pressure was measured using a Wilhelmy plate positioned in close proximity to the SPR probe. Both signals were monitored until a plateau in the SPR signal was observed (typically within 2 h).

III. MONOLAYERS AND PROTEIN ADSORPTION A.

Electrostatic Binding of Proteins to Monolayers

Proteins adsorb to floating monolayers and other interfaces to satisfy their amphipathic nature. Adsorption reduces surface energy, allowing physical bonds between protein and amphiphile to form while entropy is increased through the release of counterions [19,20]. The binding of charged molecules to oppositely charged films is a common model system since the monolayer charge density and subphase ionic strength can be tuned to increase or decrease adsorption [21–25]. However, caution must be exercised when interpreting the data since changing the ionic strength, am© 2003 by Marcel Dekker, Inc.


421

FIG. 3 Surface plasmon resonance technique adapted to the air–water interface utilizes an SPR sensor in contact with the monolayer.

phiphile ratios, or the amphiphile packing density also influences monolayer phase behavior and domain structure. B.

Amphiphile Miscibility, Phases, and Protein Adsorption

The electrostatic binding of charged proteins to oppositely charged monolayers is generally governed by the lateral charge density in the monolayer, with more charged amphiphiles leading to greater protein adsorption. However, other parameters, such as amphiphile miscibility and monolayer phase behavior, play critical roles. This was highlighted in recent studies of cytochrome c (cyt-c) adsorption to mixed monolayers of zwitterionic and anionic phospholipids by Maierhofer and Bucknall and by Ka¨sbauer and Bayler [25,26]. As expected, cyt-c adsorption increased proportionally to the fraction of anionic amphiphile in the mixed monolayer. However, when the miscibility of the two amphiphile components was systematically varied, some interesting cyt-c adsorption behavior emerged. The authors observed enhanced cyt-c binding to immiscible mixtures of the two amphiphiles (mismatched chain lengths) compared to miscible mixtures (identical chain lengths) of the same amphiphiles, but only when the films were held in the liquid condensed (LC) state. When the miscible and immiscible films were held in the liquid expanded (LE) state, they bound similar amounts of cyt-c. Clearly, local charge density and amphiphile phase differences (on the size scale of LC domains) governed protein binding, while the net amphiphile (charge) density played a lesser role. This example suggests phase behavior–dependent binding (and likely lateral distribution) of cyt-c. The axial distribution (i.e., penetration depth into the lipid) of cyt-c also depends on monolayer phase behavior and charge density. A study by Teisse demonstrated that when the monolayer packing density was decreased or the ratio of anionic to zwitterionic lipid increased, cyt-c penetrated more deeply into the films [27]. Likewise, an increase in cyt-c insertion with decreasing ionic strength was observed. These results were explained based on enhanced electrostatic inter© 2003 by Marcel Dekker, Inc.

422

Britt et al.

actions between amphiphile and protein; however, as the author pointed out, the smaller anionic phospholipid head group compared to the zwitterionic head group could also affect the interpretation of depth of insertion. Furthermore, increasing the ionic strength in order to reduce protein–lipid electrostatic binding also screens the like-charged lipids from each other, resulting in a more condensed monolayer as well as altering domain morphology [28–31]. A typical example is the calcium-induced segregation of DPPS into domains from a DPPS/DPPC mixed monolayer [29]. A final consideration regarding monolayer phase behavior and protein adsorption is protein-induced amphiphile demixing [19] and phase transitions [32]. Penetration of the protein into the monolayer reduces the area available to the amphiphiles and effectively compresses the amphiphiles. A manifestation of this process is given in Section IV, where ␲ is observed to change discontinuously upon albumin insertion into PS-PEO films. A recent review by Vollhardt and Fainerman details this remarkable event [32]. These examples should remind the reader of the subtleties associated with even simple model systems of just one or two amphiphile types and a single type of subphase protein. C.

Enhanced Protein Adsorption to Mixed Monolayers: Ferritin Binding to Mixed Cationic/Neutral Monolayers

The previous examples of the electrostatic binding of cytochrome c to oppositely charged monolayers followed the generally observed trend of increased protein adsorption with increasing fraction of charged amphiphile. In this section we show that just the opposite can happen, where decreasing the monolayer charge density by mixing a cationic amphiphile with neutral amphiphiles increases the amount of anionic protein adsorbed. The reason for this behavior will become apparent from miscibility diagrams, which reveal more positive surface potentials and effective dipole moments in the mixtures despite decreased charge densities. We have recently investigated the adsorption behavior of negatively charged horse spleen ferritin (isoelectric point 4.5) to cationic monolayers of varying charge density held at the air–water interface of the reflection trough [18]. In these experiments, the charge density was varied by two different means. In the first method, the cationic amphiphile, dioctadecyldimethylammonium bromide [(C18)2N(CH3)⫹Br⫺] (DOMA) was mixed with the neutral amphiphile, stearic acid methyl ester (C18COOCH3) (SME). These amphiphiles were prepared at various molar ratios, spread from chloroform, then rapidly compressed to 30 mN/m on preequilibrated subphases of 0.01 mg/mL ferritin in water. A second means of controlling charge density was to keep the amphiphile composition fixed but vary the packing density (surface pressure) of the amphiphiles. In this technique, single-component films of either DOMA or eicosylamine (C20NH2) were spread on ferritin subphases then compressed to surface pressures of 10, 20, or 30 mN/m. Ferritin adsorption was recorded continuously after film spreading as a change in light reflection at the interface. The ferritin binding kinetics on C20NH2 films prepared at different surface pressures are given in Fig. 4. It is apparent that the reflectivity signal increased during compression, a result of ferritin adsorption/insertion to the film before the feedback surface pressure (indicated by the arrows) was reached. As might be anticipated for an electrostatically driven adsorption, films prepared a higher surface pressures (higher charge densities) bound more protein than ‘‘loosely’’ packed films (compare © 2003 by Marcel Dekker, Inc.


423

FIG. 4 Ferritin adsorption kinetics to eicosylamine films as a function of monolayer charge density (surface pressure). The monolayers were prepared on 0.01 mg/mL ferritin containing subphases in either water (solid lines) or 150 mM PBS (dashed line). The arrows mark when the films reached the indicated surface pressures.

the plateau ⌬R values). Protein-induced changes in surface pressure were not observed above 20 mN/m, suggesting the large ferritin particles (diameter 12.5 nm) were unable to penetrate compressed films. From the ⌬R data it is noted that the rate of adsorption also increased with increasing packing density (compare the slopes preceding the plateaus). A similar trend was seen for ferritin binding to DOMA films (data not shown). These results indicate a direct correlation between monolayer charge density (controlled by surface pressure) and ferritin adsorption. However, when monolayer charge density was varied by diluting the cationic amphiphile in a neutral amphiphile matrix, an opposite trend emerged. This is seen in Fig. 5, where ferritin adsorption kinetics to a series of SME:DOMA mixtures prepared at 30 mN/m are presented. Comparing the plateau ⌬R values at 90 min a trend of increased binding with increasing SME fraction is apparent. A maximum enhancement in binding for the SME:DOMA 6:1 ratio is noted, yet even the SME:DOMA 10:1 ratio binds more ferritin than pure DOMA. This unique behavior is highlighted in the inset of the graph, where ferritin adsorption at 90 min is plotted versus mole fraction SME. A reduced rate of ferritin binding from PBS (dashed lines in Figs. 4 and 5) is indicative of the electrostatic nature of the protein–monolayer interaction. However, we should caution that a plateau was never reached from the PBS subphase, and at longer times the amount of ferritin adsorbed may very well surpass the coverages reported for adsorption from pure water [23,24]. To minimize salt-induced aggregation of ferritin (in bulk and the adsorbed state) as well as to avoid salt-driven amphiphile phase separation [28,29], salt-free subphases were favored. © 2003 by Marcel Dekker, Inc.

424

Britt et al.

FIG. 5 Ferritin adsorption kinetics to the indicated SME:DOMA mixtures held at 30 mN/m (same subphase conditions as in Fig. 4). The inset in this graph compares the amount of ferritin adsorbed at 90 min versus the mole fraction of SME in the monolayer. (Reproduced by permission of The Royal Society of Chemistry on behalf of the PCCP Owner Societies.)

From a packing density standpoint, a pure DOMA film held at 30 mN/m cor˚ 2. For DOMA mixed with responds to a charge density of about one charge per 60 A ˚ 2. Thus, SME in a 6:1 ratio, the charge density is reduced to one charge per 180 A a threefold reduction of charged amphiphile density resulted in a ⬃1.5-fold increase in bound ferritin after 90 min of adsorption (see Fig. 4). Furthermore, ferritin adsorption to SME films from water subphases was negligible, suggesting SME alone did not contribute to ferritin binding under these conditions [33]. Obviously SME acts as more than a charge diluent when mixed with DOMA. In fact, as is shown in the following miscibility analysis section, SME actually amplifies the surface potential in spite of reducing the monolayer charge density. D.

Amphiphile Miscibility and Mixing Diagrams

The unique ferritin adsorption behavior on SME:DOMA monolayers suggests that these mixtures must exhibit properties that are not a simple average of the two constituent amphiphiles. In this section we demonstrate that this is indeed the case by evaluating SME:DOMA miscibility in terms of mixing induced changes in monolayer area, surface potential, and effective dipole moment. 1. Excess Surface Potential and Ferritin Adsorption Trends A first indication of the unique SME:DOMA mixing behavior is found in the ␲-A and ⌬V-A isotherms, presented in Fig. 6. From the isotherms, it is noted that the © 2003 by Marcel Dekker, Inc.


425

FIG. 6 Surface potential– and surface pressure–area isotherms for the indicated SME:DOMA mixtures on water. (Reproduced by permission of The Royal Society of Chemistry on behalf of the PCCP Owner Societies.)

mixtures collapse at higher surface pressures (␲c) and surface potentials (⌬Vc) than the pure components. The increase in ␲c may arise from a more efficient packing ˚ 2/molec) fill in the interstitial structure in the mixture, where smaller SMEs (⬃20 A 2 ˚ /molec). The large increases in ⌬Vc spaces between the larger DOMAs (⬃60 A values for the mixtures, however, are quite unusual, since they indicate the surface potential becomes more positive (by hundreds of millivolts!) as cationic DOMA is diluted in neutral SME. In Fig. 7 miscibility diagrams constructed from the isotherms clearly illustrate the increase in ⌬V (top diagram) and improved packing density (middle diagram) with SME content. Moreover, ␮ (bottom diagram) follows the same trend as ⌬V, indicating that an increase in amphiphile dipole density (⌬Amix < 0) cannot explain the excess positive character of the mixtures since ␮ normalizes ⌬V in terms of area [recall Eq. (2)]. Other factors, such as a more orthogonal alkyl-tail dipole orientation in the mixture, changes in the electric double layer, and/or changes in the interfacial water structure must act to increase the surface potential, which in turn leads to enhanced protein adsorption. 2. Origins of Excess Properties of Mixed Monolayers If the tilt angle of the alkyl-tail dipoles in the mixture is reduced compared to the pure film, then increases in ⌬V and ␮ are expected due to the increase in the normal component of the dipoles (refer to Section II.A). In transferred films, DOMA has a sizeable tilt angle of 45⬚, which is a result of the size mismatch between the large © 2003 by Marcel Dekker, Inc.

426

Britt et al.

FIG. 7 SME:DOMA miscibility diagrams constructed from the isotherms in Fig. 6.

DOMA head group and the alkyl-tails [34]. Introducing SME into a DOMA monolayer may effectively compensate for the free space among DOMA tails, thus ordering the tails and reducing the DOMA tilt angle. Such an event has been observed for other mixed monolayers [35,36], and we have recently verified that this is also the case for SME:DOMA [37]. However, it turns out that ordering of the tails and reduction of the DOMA tilt angle in mixtures can only increase the surface potential by ⬃100 mV [18], indicating that the remaining excess potential of up to 400 mV (see Fig. 7) must arise from mixing induced changes at the head group–water interface. Interfacial water molecules spontaneously orient with the oxygen atoms directed toward the air, thus polarizing the interface and creating a potential jump on © 2003 by Marcel Dekker, Inc.


427

the order of ⫺100 to ⫺200 mV [9]. An amphiphile monolayer will depolarize this water in a manner dependent on the amphiphile head group chemistry [38,39]. Vogel and Mo¨bius proposed that amphiphiles with somewhat hydrophobic (methylated) head groups like SME disrupt the interfacial water structure and, when mixed with charged amphiphiles, increase the charge-hydration shell distance [40]. This would effectively amplify the surface potential by reducing the ability of water dipoles (and counterions) to collectively screen amphiphile dipoles and charges. Although the carbonyl dipole [41] could also contribute more strongly to the surface potential in the mixture, infrared reflectivity analysis did not reveal any changes in the SME carbonyl band upon mixing [37]. The influence of head group hydrophobicity was tested by investigating the properties of DOMA mixed with octadecanol (ODOH) as well as DOMA mixed with steric acid ethyl ester (SEE). The mixing diagrams for these films are shown in Fig. 8. Interestingly, ODOH does not lead to as compact a film as either SME or SEE (middle diagram). However, it is clearly seen that for the ODOH:DOMA mixtures ⌬␮mix remains near zero, while the SEE:DOMA mixtures, like the SME:DOMA mixtures, have excess positive ⌬␮mix values. Although ODOH:DOMA mixtures have ⌬⌬Vmix values on the order of 250 mV, recall that this can arise from increased dipole density upon mixing (⌬Amix < 0) as well as mixing induced changes in amphiphile tilt angle and alkyl-tail order. Thus, we would not anticipate enhanced ferritin binding to ODOH:DOMA monolayers as was observed for DOMA:SME monolayers. In summary, the trend of increased ferritin binding as DOMA was diluted in SME is primarily attributed to amplification of DOMA charge by SME-induced disruption of oriented water dipoles. Reduced amphiphile tilt angles and ordering of alkyl-tails also contribute to the enhanced positive character of the mixtures, albeit to a lesser extent. Combined, these effects can significantly enhance protein adsorption in a manner not expected based on a simple monolayer charge dilution model. Drawing from these insights we propose in the next section that using a mixture of PEO chain lengths in a PS-PEO monolayer can improve the protein-repellent character of the film.

IV.

PROTEIN REPELLENT MONOLAYERS

A.

PEO Surface Density Versus PEO Brush Length

Biomaterials surfaces, when modified with attached poly(ethylene oxide) (PEO) chains, display low levels of protein adsorption, especially when the surface density of PEO is sufficiently high [42–44]. This resistance to protein adsorption has been attributed to (1) steric (or osmotic) repulsion due to the compression of the PEO chains by an approaching protein molecule and/or (2) a unique property of ethylene oxide (EO) segments capable of repelling proteins. Recently, a ‘‘two-state’’ polymer model was also used to explain PEO interactions with water [45] and with dissolved proteins [46,47]. How effective the end-attached PEO chains are in preventing proteins from adsorbing to the underlying surface depends on both the length of the PEO chains and their surface density [48–51]. These two parameters are interrelated as the increase of surface density of PEO also stretches the individual chains and extends their length. For small globular proteins the PEO surface density plays a © 2003 by Marcel Dekker, Inc.

428

Britt et al.

FIG. 8 Miscibility diagrams (at 30 mN/m) of DOMA mixed with ODOH (squares), SME (triangles), and SEE (circles).

more important role than the length of PEO chains, while for large proteins or particles the length of PEO chains is the key element to control. In reality, however, it is rather difficult to fully control both the PEO surface density and the brush length as the majority of PEO surface immobilization processes are self-limited by the same effect they aim to achieve: steric self-repulsion. Because of that, there is a considerable interest to design experiments in which both the control of a wider range of densities of surface-attached PEO chains and the measurement of protein adsorption in situ without intermediary wash steps—that could remove weakly adsorbed molecules—are available. © 2003 by Marcel Dekker, Inc.


B.

429

In Situ SPR Measurements of PS-PEO Monolayer Protein Interactions

We have recently conducted a set of in situ protein adsorption experiments on surface-attached PEO chains using insoluble poly(styrene)–poly(ethylene oxide) (PSPEO) block copolymer monolayers at the air–buffer solution interface [52]. Monolayers were designed to cover a wide range of PEO surface densities, from isolated PEO chains to the packed PEO chains in the so-called brush configuration. Protein adsorption onto such monolayers was measured using an in situ surface plasmon resonance (SPR) method adapted to work at the air–solution interface, as outlined in Section II.D. By depositing varying amounts of the PS-PEO block copolymer (MWPS 12.2 kDa, 117 monomers; MWPEO 23.9 kDa, 543 monomers, Polymer Source Inc., Canada) at the interface and by keeping the predetermined barrier position constant, it was possible to set the surface pressure in the monolayer to a desired magnitude and thus achieve different surface density of PS-PEO molecules. Protein adsorption experiments were conducted at several surface pressures, ␲PS-PEO, that resulted in different PEO surface density, ␴PEO, and different PEO chain configurations (Table 1) [53,54]. Figure 9 shows the results of human serum albumin (HSA) adsorption kinetics measured with the SPR probe (RIU change versus time, upper panel) and parallel

TABLE 1 PS-PEO Monolayer Surface Pressure, the Resulting PEO Surface Density, and a Description of PEO Chain Configuration Surface pressure, ␲PS-PEO (mN/m)

PEO surface density, ␴PEO (nm⫺2)

1.6

0.00645

5.3

0.01287

9.7

0.01934

10 10.5 11

0.03245 0.05155 0.09747

12.7

0.11947

27

0.14286


PEO chain configuration (after Refs.) Very low surface density regime. PEO chains are individually adsorbed at air–buffer interface. Low surface density regime. PEO chains adsorbed at the air–buffer interface start to force each other to submerge into the buffer subphase. Intermediate surface density regime. Pseudo– first order phase transition conditions where the PEO chains are confined to a smaller and smaller area, thereby forcing other neighboring chains to leave the air–water interface and extend into the buffer subphase. Onset of the high surface density regime. PEO chains are desorbed from the air–buffer interface and all become extended in the subphase; the onset of a ‘‘brush’’ configuration. High surface density regime. Brush configuration. High surface density. PEO chains are even more compressed and stretched in the brush configuration.

430

Britt et al.

FIG. 9 HSA adsorption kinetics measured with an SPR probe (upper panel) and parallel surface pressure changes measured using the Wilhelmy plate (lower panel). The inset in the upper panel shows the control HSA adsorption kinetics onto the unmodified and hydrophobically modified SPR probe gold surfaces.

surface pressure changes measured using the Wilhelmy plate (surface pressure versus time, lower panel). The SPR method is a surface-sensitive technique with a sensing depth of a fraction of a micrometer [55]; hence, the SPR signal is proportional to the mass of protein accumulated below (i.e., adsorbed) or inserted into the PS-PEO monolayer. One finds the HSA adsorption kinetics curves distributed between two distinct groups: one group below and other above ␲ ⬇ 11 mN/m in the PS-PEO monolayer (Fig. 9, upper panel). Knowing that the surface pressure in the PS-PEO monolayer is also determining the surface density of PEO chains, one concludes that there is a threshold PEO surface density above which HSA will neither adsorb onto nor insert into the monolayer. Indeed, for ␴PEO > 0.0975 PEO chains nm⫺2 (i.e., one chain per 10.26 nm2), there is only negligible adsorption/insertion of HSA. By comparing this area per PEO chain with the radius of gyration of an isolated PEO chain © 2003 by Marcel Dekker, Inc.


431

in water (Rg = 7.67 nm) [56], one finds that at the threshold surface density the PEO chains are compressed against each other and extended in solution. C.

Protein–PS-PEO Monolayer Interactions Sensed by the Surface Pressure Changes

It is interesting that in the regime of lower PEO surface density the steady-state adsorption kinetics of HSA detected by SPR are grouped at a similar adsorption level; i.e., there is only a weak dependence of HSA adsorption on PEO surface density in the low surface density regime. At these conditions, HSA molecules are able to insert themselves into the monolayer and/or adsorb onto it. The surface pressure changes give more insight into what happens to HSA molecules. In the low surface density regime all surface pressure kinetics display a kink after the initial rise of surface pressure (solid line kinetics, lower panel in Fig. 9). The interpretation of such a kink in not straightforward as the surface pressure change is not proportional to the amount of HSA adsorbed or inserted into the monolayer. It is speculated here that the initial rise of the surface pressure up to the kink corresponds to the insertion of the HSA molecules into voids of the PS-PEO monolayer, while the changes of the surface pressure beyond each kink correspond to protein molecules unfolding at the interface. This speculation is supported by two findings: (1) each kink in surface pressure kinetics occurs at a very similar surface pressure, roughly around ␲ ⬇ 10 mN/m, and (2) the total change of the surface pressure from initial surface pressure until the step (kink) becomes smaller with the increase of the initial surface pressure. As the HSA molecules adsorb to the interface, they increase the surface pressure in the monolayer and thus force the PEO to stretch into the subphase. At the adsorption time corresponding to the kink, the SPR kinetics level off, approaching the steady state (Fig. 9, upper panel). The SPR signal will not increase past that stage, yet the surface pressure continues to rise due to the protein. When the process of HSA insertion starts at higher initial surface pressures there is less area available for the insertion of HSA into the monolayer. Finally, at high initial surface pressures, ␲ > 10.5 mN/m, the kink disappears altogether and is replaced by an induction period (dashed line kinetics, Fig. 9, lower panel). The absence of any surface pressure change in the first 30 min can be taken as another sign that the PSPEO layer is resisting protein insertion. However, at longer times a smaller increase of surface pressure typically occurs, suggesting that, although there is only very small adsorption measured by SPR, even such a small amount of adsorbed protein can lead to a surface pressure increase. Figure 10 shows the average HSA layer thickness, calculated from the SPR results, plotted as a function of PEO surface density. The layer thickness is calculated from the magnitude of the RIU changes at the steady state in each adsorption kinetics [55]. This film thickness should not be interpreted as a physical thickness of a continuous-protein thin film attached to the PS-PEO monolayer, but as an optically measured increase of the HSA mass at the interface. One important piece of information from Fig. 10 is that there is a threshold PEO surface density above which protein does not adsorb to the interface. One expects that the threshold PEO surface density is protein size dependent. That is, small proteins may penetrate the grafted PEO chains more easily than very large protein molecules; hence, a higher threshold PEO surface density may be required for preventing small size protein adsorption. © 2003 by Marcel Dekker, Inc.

432

Britt et al.

FIG. 10 Average HSA layer thickness calculated from the SPR results and plotted as a function of the PEO surface density. The layer thickness is calculated from the magnitude of the RIU changes at the steady-state level of adsorption kinetics.

Experiments with a large spherical protein, ferritin (MW 680 kDa, data not shown), indicated that the PEO surface density threshold between adsorption and no adsorption shifts to a lower value (G. Jogikalmath and V. Hlady, in preparation). In order to make a surface resistant to adsorption of different proteins from a multiprotein mixture such as blood plasma, one may explore the so-called bimodal PEO brush. In a bimodal PEO brush, the shorter PEO chains are interdispersed between longer PEO chains and expected to force the long PEO chains to extend even further from the interface [57,58]. In such a case, a dual protein repulsion action is expected: (1) large proteins are intercepted by extended long PEO chains, and (2) smaller proteins, which may penetrate in between the long PEO molecules, are intercepted by the interdispersed shorter PEO chains. Preliminary results in our laboratory indicate that the bimodal PEO brush layers are indeed more effective in preventing protein adsorption than either of the two monomodal parent PEO brush layers at the same surface density.

ACKNOWLEDGMENTS The authors wish to acknowledge financial support from NIH R01-HL44538 (for the PS-PEO protein binding study) and thank Dr. X. Du for confirming the reported ferritin adsorption behavior using the SPR technique.

REFERENCES 1. 2.

D. Mo¨bius and R. Miller, eds. Proteins at Liquid Interfaces. Elsevier, New York, 1998, pp. 1–498. A. Watts, ed. Protein–Lipid Interactions. Elsevier, New York, 1993, pp. 1–379.


Protein–Monolayer Interactions 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39.

433

Interactions of biopolymers with model membranes. International Bunsen Discussion Meeting, Halle, Germany, 2000. C. H. Giles, S. D. Forrester, and G. G. Roberts. In Langmuir–Blodgett Films (G. G. Roberts, ed.). Plenum Press, New York, 1990, pp. 1–15. H. Brockman. Current Opin. Struct. Biol. 9:438–443, 1999. H. T. Tien and A. L. Ottova. J. Membrane Sci. 4886:1–35, 2001. H. T. Tien. Bilayer Lipid Membranes (BLM): Theory and Practice. Marcel Dekker, New York, 1974. E. Sackmann. Science 271:43–48, 1996. V. Vogel and D. Mo¨bius. Thin Solid Films 159:73–81, 1988. H. Brockman. Chem. Phys. Lipids 73:57–79, 1994. R. Maget-Dana. Biochim. Biophys. Acta 1462:109–140, 1999. J. H. Fendler. Membrane-Mimetic Approach to Advanced Materials. Springer-Verlag, Berlin, 1994, pp. 1–203. G. L., Jr., Gaines. Insoluble Monolayers at Liquid–Gas Interfaces. Wiley-Interscience, New York, 1966, pp. 1–198. W. Cordroch. Polarisation unpolarer Moleku¨le in Monofilmen und Beeinflussung chemischer Gleichgewichte durch geplante Dipolschichten, in Mathematisch-Naturwissenschaft. Ph.D. dissertation, Georg-August-Universita¨t, Go¨ttingen, 1991. D. Mo¨bius, W. Cordroch, R. Loschek, L. F. Chi, D. Dhathathreyan, and V. Vogel. Thin Solid Films 178:53–60, 1989. F. MacRitchie. J. Colloid Interface Sci. 57:393–397, 1976. D. Mo¨bius. In Langmuir–Blodgett Films (G. Roberts, ed.). Plenum Press, New York, 1990, pp. 223–272. D. W. Britt, D. Mo¨bius, and V. Hlady. Phys. Chem. Chem. Phys. 2:4594–4599, 2000. S. May, D. Harries, and A. Ben-Shaul. Biophys. J. 79:1747–1760, 2000. K. Wagner, D. Harries, S. May, V. Kahl, J. O. Ra¨dler, and A. Ben-Shaul. Langmuir 16: 303–306, 2000. N Ben-Tal, B. Honig, C. Miller, and S. McLaughlin. Biophys. J. 73:1717–1727, 1997. K. S. Mayya and M. Sastry. Langmuir 14:74–78, 1998. F. Hoö¨k, M. Rodahl, P. Brzezinski, and B. Kasemo. J. Colloid Interface Sci. 208:63– 67, 1998. C. A. Johnson, Y. Yuan, and A. M. Lenhoff. J. Colloid Interface Sci. 223:261–272, 2000. M. Ka¨sbauer and T. M. Bayerl. Biochemistry 38:15258–15263, 1999. A. P. Maierhofer, D. G. Bucknall, and T. M. Bayerl. Biophys. J. 79:1428–1437, 2000. J. Teissie. Biochemistry 20:1554–1560, 1981. Y. Yuan and A. M. Lenhoff. Langmuir 15:3021–3025, 1999. M. Ross, C. Steinem, H.-J. Galla, and A. Janshoff. Langmuir 17:2437–2445, 2001. W. Knoll, G. Schmidt, H. Ro¨tzer, T. Henkel, W. Pfeiffer, E. Sackmann, S. Mittler-Neher, and J. Spinke. Chem. Phys. Lipids 57:363–374, 1991. J. Marra. J. Phys. Chem. 90:2145–2150, 1986. D. Vollhardt and V. B. Fainerman. Adv. Colloid Interface Sci. 86:103–151, 2000. X. Du, D. W. Britt, and V. Hlady. Unpublished results. K. Okuyama, Y. Soboi, N. Iijima, K. Hirabayashi, T. Kunitake, and T. Kajiyama. Bull. Chem. Soc. Jpn. 61:1485–1490, 1988. B. Casson and C. D. Bain. J. Phys. Chem. B. 103:4678–4686, 1999. G. Brezesinski, M. Thoma, B. Struth, and H. Mo¨hwald. J. Phys. Chem. 100:3126–3130, 1996. D. W. Britt, C. Selle, D. Mo¨bius, and M. Lo¨sche. (Manuscript in preparation.) J. Kim, G. Kim, and P. Cremer. Langmuir 17:7255–7260, 2001. M. Colic and J. D. Miller. In Interfacial Dynamics (N. Kallay, ed.). Marcel Dekker, New York, 2000, pp. 35–82.


434 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58.

Britt et al. V. Vogel and D. Mo¨bius. J. Colloid Interface Sci. 126:408–420, 1988. S. Diaz, F. Amalfa, A. C. Biondi de Lopez, and E. A. Disalvo. Langmuir 15:5179–5182, 1999. J. M. Harris, ed. Poly(Ethylene Glycol) Chemistry Biotechnical and Biomedical Applications. Plenum Press, New York, 1992. D. L. Elbert and J. A. Hubbell. Annu. Rev. Mater. Sci. 26:365–394, 1996. J. D. Andrade, V. Hlady, and S. I. Jeon. Adv. Chem. Ser. 248:51–59, 1996. R. L. C. Wang, H. J. Kreuzer, and M. Grunze. J. Phys. Chem. B 101:9767–9773, 1997. A. Halperin and D. E. Leckband. C. R. Acad. Sci. Paris t1 IV:1171–1178, 2000. D. E. Leckband, S. Sheth, and A. Halperin. J. Biomater. Sci. Polym. Ed. 10:1125–1147, 1999. S. I. Jeon and J. D. Andrade. J. Colloid Interface Sci. 142:159–166, 1991. S. I. Jeon, J. H. Lee, J. D. Andrade, and P. G. de Gennes. J. Colloid Interface Sci. 142: 149–158, 1991. S. J. Sofia, V. Premnath, and E. W. Merrill. Macromolecules 31:5059–5070, 1998. A. Halperin. Langmuir 15:2525–2533, 1999. G. Jogikalmath. Ph.D. dissertation, University of Utah, Salt Lake City, UT, 2002. H. D. Bijsterbosch, V. O. de Haan, A. W. de Graaf, M. Mellema, F. A. M. Leermakers, M. A. Cohen Stuart, and A. A. van Well. Langmuir 11:4467–4473, 1995. M. C. Faure, P. Bassereau, M. A. Carignano, I. Szleifer, Y. Gallot, and D. Andelman. Eur. Phys. J. B(3):365–375, 1998. L. S. Jung, C. T. Campbell, T. M. Chinowsky, M. N. Mar, and S. S. Yee. Langmuir 14: 5636–5648, 1998. K. Devanand and J. C. Selser. Macromolecules 24:5943–5947, 1991. A. M. Skvortsov, A. A. Gorbunov, F. A. M. Leermakers, and G. J. Fleer. Macromolecules 32:2004–2015, 1999. E. P. K. Currie, M. Wagemaker, M. A. Cohen Stuart, and A. A. van Well. Macromolecules 32:9041–9050, 1999.


17 Local and Global Optical Spectroscopic Probes of Adsorbed Proteins VLADIMIR HLADY JOS BUIJS

I.

University of Utah, Salt Lake City, Utah, U.S.A.

Uppsala University, Uppsala, Sweden

INTRODUCTION

The conformation of adsorbed proteins has a significant influence on the bioactivity of the proteins or subsequent processes like biofouling, implant rejection, and others. The design of protein-contacting materials requires a detailed knowledge of protein conformation in the adsorbed state as a function of the physical and chemical properties of the sorbent surface. In addition to the protein biological function, it has been shown that conformational changes are a crucial factor in describing the adsorption process. Nowadays, there are many experimental methods available to directly access information on the conformation of adsorbed proteins. Among them, one of the most prominent techniques is optical spectroscopy. In this chapter we review the principles of optical spectroscopy as they apply to a population of protein molecules adsorbed at an interface. We start by reviewing the methods for enhancing the signal from the interfacial molecules and follow with a short description of three spectroscopic techniques most often used in protein adsorption studies: fluorescence, infrared absorbance, and circular dichroism spectroscopy.*

A.

Interaction of Protein Molecules with Light

The interaction between light and molecules represents one of the fundamental problems in quantum optics [4]. It is impossible to visualize a single light quantum, a photon, propagating at the speed of light through space because of the inherent uncertainty about the photon’s position. Nevertheless, one can imagine a stream of photons moving through space: a light beam. The electromagnetic field associated

*This choice is based solely on the authors’ experience and preferences rather than on the applicability of spectroscopic techniques to protein adsorption studies or on the quality of spectroscopic information about adsorbed proteins. The reader is referred to recent monographs [1–3] which describe other optical spectroscopic techniques such as surface enhanced Raman scattering, second harmonics generation, and others.


436

Hlady and Buijs

with each proton will induce a time-dependent charge redistribution of a finite dimension in its surrounding. If the space through which the stream of photons moves is occupied by a protein molecule, the entities interacting with the photons’ electromagnetic field may be a covalent bond between two atoms, a molecular orbital, some delocalized states, or some vibrational or rotational modes involving a part or whole protein molecule. The complexity of these interactions is simplified by modeling the molecule as a collection of dipoles which interact with the propagating electromagnetic field [5]. It will take the electromagnetic field traveling at the speed of light a mere 3 ⫻ 10⫺18 s to zip through a protein of average size. And yet the effect of the traveling electromagnetic field on the surrounding dipoles in the molecule produces experimentally measurable quantities: changes in transmitted light intensity, polarization, and rotation and differences of energy and angle of the scattered light. Measuring and interpreting these changes falls in the realm of optical spectroscopy. B.

Surface Versus Bulk Sensitivity

The ultimate success of every experimental technique depends on how well the population of molecules of interest is represented in the experimental signal. Translated to the language of protein adsorption: how easy is it to distinguish the signal related to the adsorbed protein population from the signal originating from everything else in the system? Here, it is illustrative to compare the protein number density per unit volume with the number density per unit area. As an example, let’s take a 0.1cm-thick flow cell interfaced at two opposing sides by two optically transparent, flat 1-cm2 area surfaces. If a 1-mg/mL protein solution flows through such a cell, adsorption to these surfaces may develop, leading to a fraction of ␮g/cm2 of adsorbed proteins. We assume that the adsorbed amount is a hefty 0.4 ␮g/cm and find the protein bulk-to-interface ratio to be above 100. When such a cell is placed into a spectrometer and analyzed in transmission mode, less than 1% of the signal will originate from adsorbed molecules: the signal will be dominated by the bulk solution protein molecules. Clearly, a method is needed for suppressing the signal from the bulk dissolved protein molecules, while enhancing the signal from the interfacial protein population. The suppression of the bulk signal can be achieved by making the cell much thinner than 0.1 cm, which, at some point, becomes quite impractical. The second approach, namely, to enhance the signal from adsorbed molecules, can be achieved either by enhancing the electromagnetic field right at the interface and/ or by confining it to the interfacial region only. Nearly all surface-sensitive spectroscopic techniques use some form of electromagnetic field enhancement; some by using an evanescent surface wave, others by using highly reflective and metal surfaces. 1. Optics of Total Internal Reflection and Evanescent Surface Wave A light beam propagating through a solid and encountering a solid/liquid interface will totally reflect from it when the angle of incidence becomes larger than the ‘‘critical angle,’’ ␪c:

␪c = sin⫺1

冉冊 n1 n3

(1)

where n1 and n3 are the refractive indices of the liquid and the solid, respectively. © 2003 by Marcel Dekker, Inc.

Spectroscopic Probes of Adsorbed Proteins

437

Even at the total internal reflection conditions, some of the electromagnetic field penetrates the interface and delivers the electromagnetic energy to the interfacial region in the liquid by propagating parallel to the surface by a Goos–Hansen shift. The theoretical basis of evanescent wave optics has been well described [6–8] and only a simple description of the evanescent wave will be given here. The electric field amplitude of the evanescent wave, E, decays exponentially in solution with distance normal to the interface (z-direction): E = E0e⫺z/d

(2)

where E0 is the electric field amplitude right at the interface and the characteristic penetration depth, d, is defined by the angle of incidence, ␪, wavelength of light in free space, ␭0, and refractive indices of the solid and the liquid (Fig. 1a): d = ␭0(2␲)⫺1(n32 sin2␪ ⫺ n12)⫺1/2

(3)

The magnitude of d is on the order of ␭0 or smaller except at the critical angle conditions where d → ⬁. Hence, the molecules in the region outside of the evanescent surface wave (at z >> d) will not feel the presence of the electromagnetic field. Moreover, the intensity of the evanescent surface wave, Ie, is proportional to E 2, and increases to nearly twice the incident light intensity as one approaches the surface. This enhancement of the intensity combined with the exponentially decaying field improves the surface sensitivity of spectroscopic methods which utilize the evanescent surface wave. At low surface coverage the presence of protein molecules at the interface will not affect the optics of total internal reflection (TIR). However, when a dense protein layer is formed, it will present itself to the electromagnetic field as another optically distinct layer with a refractive index, n2, affecting the local electric field amplitude. The details on the TIR optics of three- and multilayered interfaces can be found elsewhere [8–10]. 2.

Surface Selection Rules and Metal Surface Enhancement of the Electromagnetic Field

Smooth metal surfaces reflect light very efficiently. In the external reflection from metal surfaces at the near-grazing incidence angle, light experiences a 90⬚ phase shift upon the reflection. A shift in phase by 90⬚ for the light polarized perpendicularly to the metal surface will nearly double the electric field amplitude at the site of reflection, while the electric field amplitude of the light polarized parallel to the metal surface will cancel itself upon reflection. The process results in an electric field enhancement for only one light polarization: the reflection at the metal surface produces a surface selection rule. Only those molecules whose transition dipoles are oriented perpendicularly to the surface will effectively interact with the incoming light (Fig. 1b) [5]. The advantages of external reflection spectroscopy are based on this surface selection rule; the main disadvantage is the presence of the electromagnetic field in the bulk phase outside of the metal, which precludes the use of this method in situations where dissolved protein molecules are present in the bulk phase. Some metal surfaces, like metallic silver, will, when roughened on the scale of tens of nanometers, enhance the electromagnetic field in their very close proximity. The enhancement is quite large (up to a factor of 106!) and decays rapidly at very small distances away from the metal surface (Fig. 1c). This electromagnetic field © 2003 by Marcel Dekker, Inc.

438

Hlady and Buijs

FIG. 1 Three mechanisms of enhancement of the electromagnetic field at interfaces. (a) The optics of total internal reflection creates an evanescent surface wave which decays on the optically rare side of the interface with a characteristics length of dp. (b) The phase shift upon the reflection of light at a smooth metal surface cancels the electric field components (Es) oriented perpendicular to the plane of incidence while nearly doubling the components oriented parallel to the plane (Ep). The resulting electric field at the site of reflection (E Rp ) will preferentially interact with the molecular transition dipoles that are oriented in the same way (i.e., perpendicular to the surface). (c) Roughened metal surfaces or small metal islands will locally enhance the strength of the incident electromagnetic field. The enhanced field will, in turn, interact with the molecules that are in contact with the metal. The enhancement effect i i l di f di R i d fl



439

enhancement is frequently utilized in surface enhanced Raman scattering (SERS) experiments [11]. The advantage of such a large electromagnetic field enhancement is found in compensation for the high number/density ratio between the bulk and adsorbed proteins. Because of the use of a metal surface, SERS is often combined with some electrochemical means of studying protein adsorption [12,13]. The main disadvantage is the need for a roughened metal surface, although a thin overlayer of a dielectric has been shown to leave enough of the enhanced electromagnetic field at the dielectric/air (or liquid) interfaces [5]. The choice is now for the researcher to decide which enhancement mechanism to use (Fig. 1a–c): (1) evanescent surface wave of given energy (␭), (2) external reflection at a smooth metal surface and its surface selection rule, and/or (3) a roughened metal surface enhancement. These three types of enhancements will not all be applicable to every spectroscopic method. For example, fluorescence emission is often quenched near metal surfaces, and a dielectric overlayer is required to displace the protein layer away from the metal to avoid quenching. A roughened metal surface will not support specular reflection and the reflectance/absorbance measurements can be difficult. Furthermore, the optical requirements for the sorbent surfaces will limit the variety of sorbent materials suitable to study protein adsorption. Similar to the optical spectroscopy of proteins in the bulk phases, the use of a single spectroscopic technique is less likely to succeed in probing different responses of adsorbed protein molecules. A more comprehensive, multitechnique approach is always preferred and recommended.

C.

Proteins’ Private and Public Lives

Which information about the protein structure and dynamics is experimentally accessible to optical spectroscopy measurements? Optical spectroscopy may seem to be a poor choice for studying protein structure when compared with x-ray diffraction and NMR spectroscopy. Yet, these two techniques require milligrams of protein, either in an ordered, crystalline state (x-ray diffraction) or in solution (NMR) and will both provide only time-average structural information [14]. Although optical spectroscopic techniques do not give comprehensive structural information, they do require much less protein and provide structure-related information in a dynamic, time-resolved fashion that depends on the duration of a particular spectroscopic event. In terms of a molecule’s dynamics we differentiate between a protein’s ‘‘private’’ and ‘‘public’’ life [15]. In a protein’s private life we include the events that are occurring on picosecond time scales; in other words, the events that are short enough to involve the protein itself or allow only a very limited coupling with the protein’s surrounding. Absorption of a photon by a protein chromophore or excitation of various vibrational and rotational modes of the molecule are examples of a protein’s private life. Other spectroscopic events may occur on time scales of a nanosecond and longer, which will allow enough time for the protein’s public life, such as ligand binding, diffusion to/from surfaces, and other even slower events, to ‘‘show’’ through. For example, the lifetime of a fluorophore in a protein is on the order of a few nanoseconds, a sufficient amount of time for fluorescence spectroscopy to sample changes in the dielectric environment around the fluorophore or monitor © 2003 by Marcel Dekker, Inc.

440

Hlady and Buijs

how the local (fluorophore) and global (protein) rotation influence the polarization of the emitted light.

D.

Local Versus Global Spectroscopic Probes of Protein Structure and Dynamics

The interaction of light with a protein molecule may occur locally at the level of few covalent bonds, as well as more globally when a larger set of covalent bonds interacts with the light’s electromagnetic field in a collective fashion. Each observed spectroscopic phenomenon thus contains information about its locus of origin. In order to simplify the analysis of spectroscopic observations, we divide them into two large groups: local and global spectroscopic probes.* The techniques which primarily provide local spectroscopic information about a protein molecule are fluorescence and SERS spectroscopy. Absorption of light, especially in infrared and circular dichroism spectroscopy will be considered here as global spectroscopic probes because they sample and sum multiple local interactions. Figure 2 shows a protein structure of human growth hormone (hGH) with the various loci of global and local spectroscopic information. For example, a single tryptophan residue in the protein molecule (Fig. 2a) will be the locus of fluorescence emission in UV. Alternatively, an externally added fluorophore, like dansyl or some other, may also be used to obtain local spectroscopic information. Global techniques, like circular dichroism (CD) operated in the far UV, will selectively interact with optically active moieties in the protein molecule, such as the ␣-carbon atoms in the protein’s polypeptide backbone. The secondary structures in which these atoms participate, such as ␣-helix, ␤-sheet, or random coil, will produce differences in the CD spectra (Fig. 2b). CD sampling of other asymmetric moieties in the protein molecule will occur when the light wavelength coincides with the moiety absorption band. The SERS spectroscopy may locally sample the polypeptide backbone and amino acid side chains in one of the ␣-helices (Fig. 2b). Infrared (IR) spectroscopy will sample IR-active vibrational modes of a protein’s covalent bonds. In some cases these modes are affected by the protein secondary structure. For example, the C — O and N — H groups of the peptide bond form hydrogen bonds whose arrangement depends on the particular type of protein secondary structure (Fig. 2c). The information about the protein secondary structure is typically sought in the absorption bands of a protein’s amide groups [16]. Both global and local spectroscopic probe techniques will provide valuable information about protein structure and dynamics. In applying these techniques to the adsorbed protein molecules, one is interested in the changes of protein structure and dynamics caused by the process of adsorption to a particular interface. Hence, one needs a reference state. In most cases, the reference protein state is its native conformation in bulk solution, although an adsorbed protein at some ‘‘standard’’ interface can also be designated as a reference state.

*We keep in mind that such a sharp distinction between the two types of protein probes may not be entirely justified in all cases.



441

FIG. 2a, b Examples of local and global optical spectroscopic probes in a protein molecule (human growth hormone, hGH). (a) Single Trp residue (Trp86) in hGH molecule acts as a local fluorescent probe. (b) Ribbon drawn through the hGH polypeptide backbone indicates various secondary structural elements which affect the hGH CD spectrum in the far UV.


442

Hlady and Buijs

FIG. 2c Ball-and-stick model of the hGH polypeptide backbone containing the C — —O and N — H groups used as probes in IR spectroscopy.

II.

FLUORESCENCE AS A LOCAL SPECTROSCOPIC PROBE OF PROTEIN MOLECULES

The success of fluorescence spectroscopy in studying dynamics of biological macromolecules lies in its ability to access local fluorophores in biological macromolecules and use the lifetime of their excited states as a time-window for the observation of molecule dynamics [17,18]. In particular, the nanosecond and longer fluorescence time-window enables one to view the evidence of a protein’s ‘‘public’’ life, i.e., its interactions with the surrounding solvent molecules, ligands, substrates, and interfaces. In the following section we will review the essential features of various fluorescence probes as they apply to the adsorbed protein structure and dynamics. A.

Protein Fluorescence

Nearly all proteins fluoresce naturally. The intrinsic protein fluorescence has been the subject of several monographs [17–19]. The presence of tryptophan (Trp), tyrosine (Tyr), and phenylalanine (Phe) residues in protein structure is responsible for intrinsic protein fluorescence in UV. In some proteins, prosthetic groups, like pyridoxyl phosphate in phosphorylase B or the heme group in cytochrome c, are also fluorescent. In UV, the fluorescence from Phe is so weak that its contribution can be safely neglected. The protein absorption band at 280 nm is due to Trp and Tyr residues. Between the two, the Trp residue has a larger extinction coefficient; however, there are usually more Tyr than Trp residues in proteins. The two also partic© 2003 by Marcel Dekker, Inc.


443

ipate in an energy transfer process which may complicate the interpretation of observed fluorescence. In order to avoid problems of Tyr → Trp energy transfer, one often excites protein UV fluorescence at 295 nm, which will predominantly excite Trp residues. The most commonly used feature of protein fluorescence originates from the Trp residues, whose emission is a sensitive probe of the polarity of each residue’s environment and its dipolar relaxation. Several spectral forms of protein Trp residue fluorescence have been categorized and experimentally observed in proteins (Table 1) [19]. The spectral form A has been found to exist only in two proteins: azurin and bacteriorhodopsin. The spectral form S originates from Trp forming a 1:1 excited state complex (exciplex) with a polar group inside a hydrophobic protein core. The form I is from a Trp residue which forms a 2:1 exciplex with some polar group. The larger Stokes shifts of the spectral forms II and III originate from Trp residues in contact with water molecules. The difference between the two forms comes from the mobility of water molecules in contact with Trp: in the form II the Trp indol moiety is in contact with water molecules whose mobility is much smaller than the mobility of the bulk water. In the spectral form III the Trp residue is in contact with the bulk water molecules. Protein unfolding usually leads to a full exposure of Trp residues and the emergence of the Trp spectral form III. The changes of fluorescence emission maximum have to be carefully considered when interpreting the intrinsic fluorescence from adsorbed proteins. For example, a red shift in adsorbed protein fluorescence emission (i.e., a shift toward longer wavelengths) can be incorrectly interpreted as a (partial) protein unfolding occurring upon adsorption. As stated above, it has been shown that the Trp residue reports on the local changes in its environment as sampled during the lifetime of its excited state [20]. The absorption of a photon by the indole ring in Trp residue instantaneously changes its transition dipole, which, in turn, forces all other dipoles in the vicinity of the residue to respond. If the induced dipolar relaxation is fast, on the time scale of the excited state, the transition dipole–induced dipole relaxation will take some energy away from the excited state and thus shift the energy of emitted photons toward the longer wavelengths. On the contrary, if the induced dipolar relaxation is slow on the time scale of the excited state, the transition dipole will cease to exist before the dipolar relaxation will be able to take energy from the excited state, and thus no wavelength shift will occur. In these two and other similar

TABLE 1 Emission Maxima and Bandwidths in Different Spectral Forms of Proteins’ Trp Fluorescence Spectral form A S I II III a

␭cm maximum (nm)

Bandwidth (nm)

307 (sp300)a 316–317 (sh305–307, sh320–330)a 330–332 340–342 350–353

— —

sp: secondary peak; sh: shoulder.


48–50 53–55 59–61

444

Hlady and Buijs

scenarios,* the protein’s global structure does not have to change in order for the spectral changes to be observed: the only change needed is in the very local environment of the excited state. The rate of local dipolar relaxation may be affected by the presence of an interface and various surface forces. For example, the interface to which protein molecules adsorb may be charged. The electrostatic field of a charged interface will affect both the orientation of the protein’s dipoles and the rates by which they will respond to the newly created transition dipole moment of the excited state. The excited-state lifetime is sometimes taken as a measure of protein unfolding in solution [17]. However, in the interfacial protein studies it is important to recognize that the lifetime of any excited state is also affected by the proximity of a dielectric interface [21–23]. Accordingly, one expects the process of proteins’ adsorption to result in a changed fluorescence quantum yield, as noted in several studies [24–27]. 1.

Extrinsic Fluorophores

In the experimental protein adsorption studies it is frequently convenient or necessary to add an external fluorophore to the protein molecule. Such a fluorophore may have some experimentally advantageous spectral property, like an adsorption band which coincides with the wavelength of laser lines, a large Stokes shift, a high fluorescence quantum yield, or a high sensitivity to its polarity. Typically, two kinds of extrinsic fluorescent probes for protein studies are distinguished: noncovalent probes which bind to protein molecules as ligands in a reversible fashion, and covalent probes that can be bound to protein molecules in an essentially irreversible fashion through a covalent bond. The various properties of the external fluorescent probes have been summarized in the literature [28–30]. Covalently attached fluorescent probes are nowadays commonly used in fluorescence microscopy for the visualization of protein spatial distribution within the sample [31]. Chemically patterned adsorption substrates will often display differences in adsorbed protein density that can be visualized by fluorescence microscopy [32]. Many concerns about the fluorescence emitted by species associated with interfaces already made above apply to extrinsic protein fluorophores as well. Although different fluorescent labels have been frequently used in protein adsorption studies, there is always a concern about the differences in the interfacial activity between labeled and nonlabeled proteins. In addition, a noncovalently bound probe may have a different affinity for the adsorbed protein than for the same protein in solution [33]. On the other hand, the sensitivity of some extrinsic probes to the polarity of their local

*Several scenarios are possible. Assume that in the reference state (i.e., protein in solution) the Trp residue excited state experiences a partial loss of the excited state energy into its dipolar environment which results in the emission of a fluorescence photon at the wavelength, ␭sol (the ‘‘partially relaxed excited state’’). Consider the first case: upon protein adsorption the Trp fluorescence lifetime becomes shorter and/or the dipolar relaxation around the excited residue becomes slower. As a result of the shorter lifetime less energy can be taken from the excited state, resulting in a blue-shifted fluorescence (␭ads < ␭sol). Here is the second case: upon protein adsorption the fluorescence lifetime becomes longer and/or the dipolar relaxation around it becomes faster. The resulting energy of the emitted photon becomes lower as seen in a red-shifted fluorescence (␭ads > ␭sol). We leave to the reader to explore other scenarios starting from (1) initially fully relaxed Trp excited state and (2) initially fully unrelaxed Trp excited state.



445

environments [34] and to the state of protein aggregation in the adsorbed layer [35] makes them an attractive choice for protein interfacial studies. 2.

Fluorescence Energy Transfer and Polarization

The fluorescence emission from an excited state is rather sensitive to the presence of other nearby molecules, as manifested by ready quenching of fluorescence by external quenchers. One of the quenching mechanisms occurs via singlet–singlet energy transfer [36] that takes place when an excited fluorophore (a ‘‘donor’’) interacts via dipole–dipole interactions with another chromophore (an ‘‘acceptor’’) and transfers its excited-state energy before it can emit a fluorescence photon and return to its ground state. Because the efficiency of energy transfer is very sensitive to the distance from the donor to the acceptor, the fluorescence energy transfer has been called a spectroscopic ‘‘ruler’’ [37] and used to experimentally measure the distances between the residues involved in the transfer. When the energy transfer pair is in an isotropic dilute medium, the efficiency of the energy transfer is related to the inverse sixth power of the separation distance between the donor and the acceptor, spectral property of each chromophore, their orientation, the refractive index of the medium between them, and the donor fluorescence quantum yield [17]. The ability to attach extrinsic fluorophores to proteins’ selected sites has resulted in a number of experimental studies in which the energy transfer has been used for measuring distances in dissolved and membrane proteins [38]. The application of the ‘‘spectroscopic ruler’’ to a population of adsorbed protein molecules is, however, complicated by the presence of a dielectric interface [21]. The process of adsorption concentrates proteins at the interface and thus brings more than one acceptor in the close proximity of a potential energy donor. The two-dimensional arrangements of donors and acceptors may result in a different energy transfer versus average distance law than observed in dilute solutions [39]. The fluorescence of the donor and the absorption by the acceptor may both be affected by the adsorption and need to be independently determined for the adsorbed molecule [40]. The polarization of fluorescence is a powerful tool in protein adsorption studies since it is affected by the flexibility, mobility, and the state of aggregation of adsorbed molecules [41]. A careful choice of extrinsic fluorophore in terms of its limiting polarization and lifetime enables one to fine tune the time window during which these dynamic properties of adsorbed molecules can be evaluated. In general, upon a polarized excitation the emitted fluorescence will be depolarized if the fluorophores gain an excess local mobility or if the adsorption results in a high local concentration of fluorophores leading to a homotransfer of the excitation energy. B.

Total Internal Reflection Fluorescence Spectroscopy

A technique which has been successfully applied to study proteins at interfaces is total internal reflection fluorescence (TIRF) spectroscopy [42,43]. The main reasons for this success are the high sensitivity of fluorescence spectroscopy in combination with the selectivity of the evanescent surface wave for excitation of proteins, which are at or close to the adsorbent surface allowing thus for an in situ monitoring of small amounts of adsorbed proteins. Current developments in both the method of excitation and the detection of the fluorescence signals made it possible to monitor © 2003 by Marcel Dekker, Inc.

446

Hlady and Buijs

the time-resolved fluorescence of a single molecule at an interface [44]. The popularity of TIRF spectroscopy also lies in its ability to monitor more than one important aspect of the protein adsorption process. The fluorescence emitted by the adsorbing proteins can be quantified and followed in real time to study adsorption and desorption kinetics. If fluorescence is emitted by an intrinsic protein fluorophore, like Trp, one can also monitor changes in the polarity of the local environment of the fluorophore and relate them to changes in protein dynamics and/or conformation. Furthermore, fluorescence polarization measurements can be performed to study the orientation and rotational mobility of adsorbed proteins, while photobleaching experiments allow one to study protein exchange processes at the interface and lateral mobility. Lateral mobility studies are discussed in Chapter 9. The basic principle of TIRF spectroscopy is the excitation of fluorophores by an exponentially decaying evanescent wave formed at an interface (Section I.B.1) and the subsequent detection of emitted fluorescence. To perform a TIRF experiment one needs the basic spectroscopic instrumentation, such as an excitation source, a photon detection system, lenses to control and focus the excitation and emission light, and a total internal reflection element. Addition of the wavelength-selective components, such as monochromators or cutoff filters, which enable accurate selection of the wavelength of the excitation and emission, are required for the spectroscopic analysis of adsorbed protein fluorescence. Higher intensities of monochromatic light for excitation can be obtained using a laser as an excitation source. The disadvantage is the limited choice of laser line wavelengths. A continuous light source such as a high-pressure Xe arc lamp in combination with a monochromator allows for a selection of the excitation wavelength and enables experiments in UV and visible spectral ranges. The charge-coupled devices’ (CCD) photon detection systems extend the TIRF capabilities to both spectral and spatial analyses of the fluorescence. Additional optical components such as polarizers can be added in the optical pathway of the excitation and emission source for polarization measurements. In designing the TIRF instrument one faces the concerns about the instrumental sensitivity which are similar to standard solution fluorescence spectroscopy [17]. At given excitation and emission wavelengths, the fluorescence emitted by molecules in the evanescent wave region is proportional to the probability of light absorption characterized by the molecular extinction coefficient, the concentration of molecules, the intensity of the evanescent wave, and the probability of emission characterized by the fluorescence quantum yield [42]. The concentration and intensity are both functions of the distance from the interface into the cell. Another parameter which determines the detected fluorescence intensity is the instrumental factor, which includes the efficiency of fluorescence collection and the instrumental sensitivity. In TIRF experiments this factor can be obtained by a calibration procedure using fluorescence standards of relatively high concentrations. Once the instrumental factor is known, the observed fluorescence signal from adsorbed proteins can be recalculated into a number proportional to the surface density of adsorbed protein molecules by the ratio between the fluorescence quantum yields of adsorbed proteins and fluorescence standards used [24]. It is, however, also possible to quantify the amount of adsorbed protein by using independent calibration methods [24,45,46]. Many studies of protein adsorption kinetics are performed using extrinsic covalent probes, like fluorescein isothiocyanate (FITC), which absorbs and emits light in the visible wavelength region and has a relatively high quantum yield. © 2003 by Marcel Dekker, Inc.


447

TIRF spectroscopy has been used in numerous protein adsorption studies, and basic principles for the study of adsorption kinetics and description of the earlier work using TIRF for both intrinsic and extrinsic fluorophores are reviewed in the literature [8,42,43]. More recently, TIRF spectroscopy has been employed to study more complex protein adsorption systems. The versatility of a CCD camera coupled to the output of a monochromator was employed to study spatially resolved protein adsorption kinetics [47] on heterogeneous surfaces, like chemical gradients [48,49] and patterned surfaces [32]. Another TIRF example is the use of energy transfer probes attached covalently to the same protein molecule. The labeled protein was allowed to adsorb onto a dichlorodimethylsilylsilica (DDS) gradient surface, and TIRF spectroscopy was used to sample the energy transfer efficiency as a function of time in a spatially resolved way (Brynda and Hlady, unpublished). Figure 3a shows the comparison between the fluorescence spectra of IgG molecules which were double labeled with fluorescein (used as a fluorescence donor, approx. 3.6 fluoresceins/ IgG) and tetramethylrhodamine (used as a fluorescence acceptor, approx. 2.7 rhodamins/IgG) recorded for the IgG in dilute solution and when it was adsorbed to a DDS surface. The fluorescence spectra recorded during the first 15 s of the doublelabeled IgG adsorption onto the DDS surface are shown in Fig. 3b. It is evident that the initially adsorbed IgG molecules undergo rapid changes that lead to a more efficient energy transfer process than the transfer in dissolved protein. In addition to the adsorption kinetics studies, other aspects of the protein adsorption process can be inferred from the fluorescence emitted by protein’s Trp residues [50]. As stated above this residue acts as a local probe of the protein’s structure with its fluorescence emission being very sensitive to the polarity and dipolar relaxation of its local surroundings [51]. Another approach is the addition of a fluorescence quencher to a population of adsorbed proteins. The quencher will adsorb the energy of the excited state of the fluorophore, thereby inhibiting the fluorescence signal. This type of energy transfer is only possible if the quencher comes in close contact with the fluorophore; hence, the quenching reveals the accessibility of a given fluorophore to the quencher. One should take into account the possibility that the concentration of the quencher near the interface can be affected by the surface, especially if both the sorbent surface and the quencher are charged [52]. Yet another way to analyze fluorescence from adsorbed proteins is to follow the decay of the excited state in adsorbed proteins [35,53,54]. Fluorescence lifetime studies reveal information on the deexcitation processes in much more detail than can be observed using the steady-state fluorescence. The time-dependent measurements, however, do require very sensitive fluorescence instrumentation and are not as easily employed to study adsorbed proteins as the steady-state fluorescence methods. 1.

Orientation and Mobility of Adsorbed Proteins Studied with TIRF

TIRF spectroscopy can give information on the rotational mobility of fluorescent moieties by monitoring the polarization of emission (anisotropy) as a function of the polarization of the excitation light. In principle, this method measures the average motion of the transition dipole during the excited-state lifetime of a fluorophore. In the TIRF geometry the measured anisotropy reflects all rotational motions relative to the planar interface, which can originate from wobbling of the fluorophore inside the protein as well as from rotation of the whole protein molecule. The analysis of © 2003 by Marcel Dekker, Inc.

FIG. 3 Energy transfer between fluorescein (donor) and tetramethylrhodamine (acceptor) attached to a single IgG molecule (F-T-IgG). The fluorophores were attached to the IgG using the isothiocyanate chemistry (approx. 3.6 fluoresceins and 2.7 rhodamins per IgG). (a) Comparison between the fluorescence spectra of F-T-IgG in dilute solute and adsorbed to a methylated silica (DDS) surface. The ratio between the fluorescence intensity of the acceptor and the donor, I(A)/I(D), was 0.4 and 1.6 for the solution and adsorbed protein, respectively. The unfolding of F-T-IgG in urea solution did not change the solution spectrum. (b) Fluorescence spectra recorded in 1-s intervals during the first 15 s of the F-T-IgG adsorption onto the DDS– silica gradient surface. The spectra were simultaneously recorded on the three distinct surfaces: on the DDS (upper panel) and silica ends (lower panel) of the gradient surface and on the silica surface, which was approximately half covered with the methyl groups (middle panel). The I(A)/I(D) ratio was found to be different for the three surfaces. It increased as a function of time in the case of the DDS surfaces indicating that rapid changes in the structure of the adsorbed IgG layer are initially taking place (Brynda and Hlady unpublished data)



449

rotational mobility from the anisotropy measurements in a two-dimensional system differs from those in a three-dimensional system as described by Morrison and Weber [41]. It is often found that the protein mobility in the adsorbed state is reduced relative to the mobility in dilute solution and can result in a negligible rotational motion during the fluorescence lifetime [50]. Orientation of adsorbed proteins can also be determined using TIRF spectroscopy. The orientation measurements are performed by varying the polarization of the excitation source, which results in a change in the direction of the electric field components of the evanescent wave. The variation in the direction of the electric field results in a selective excitation of those transition dipole moments of fluorophores which are aligned with the field. Consequently, the observed fluorescence intensity depends on the orientation of the fluorophores. The method, first used to study the orientation of dye molecules in phospholipid monolayers [55], was later employed to study the average orientation of adsorbed proteins [56]. By measuring a second variable, such as the polarization of the emitted fluorescence, an orientation distribution of fluorophores can be obtained. Polarized excitation studies of the orientation of adsorbed proteins has been carried out in several laboratories using the porphyrin group of cytochrome c as a model fluorophore [57–60]. The porphyrin group in cytochrome c has a well-defined and relatively large transition dipole moment which is more or less in a fixed position relative to the protein structure. Full information on the orientation of adsorbed proteins also requires knowledge of the distance between the protein and the surface. As the fluorescence intensity is proportional to the intensity of the evanescent wave and the fluorophore concentration, a density profile of the fluorophore can in principle be found by monitoring the fluorescence as function of the evanescent wave intensity profile. Variation of the angle of incidence [see Eq. (3)] changes the intensity profile of the evanescent surface wave in solution, thereby offering the possibility for determining the fluorophore density profile. For monolayers of adsorbed proteins, however, this method has only a limited accuracy [61] because the evanescent wave penetration depth is two orders of magnitude larger than the size of most proteins. Consequently, changing the angle of incidence results in a rather small variation in overall fluorescence intensity from an adsorbed protein layer and the interpretation of experimental data requires an extremely high accuracy of the measurement. A more recent approach is the introduction of an additional experimental variable, namely, the electrical double layer which develops at a charged interface. The electrical double layer has a sharper gradient in the z-direction than the evanescent surface wave and can be controlled by the solution’s ionic strength. It has been demonstrated that this sharper profile can be employed in a TIRF study to determine the protein orientation using fluorescein-labeled protein, where fluorescein acts as a pH-sensitive probe [62]. 2.

Adsorption-Induced Conformational Changes in hGH Monitored by Fluorescence Spectroscopy

In this section we present an example of how the fluorescence signal, emitted by a single intrinsic Trp residue in a model protein, hGH, is used to study adsorptioninduced conformational changes. For this purpose, hGH was adsorbed onto a hydrophobic methylated silica surface from a solution containing 0.1 mg/mL protein and buffered with 0.15 M PBS at pH 7. More experimental details and a similar study © 2003 by Marcel Dekker, Inc.

450

Hlady and Buijs

of the hGH adsorption onto a variety of surfaces, including self-assembled monolayers and Langmuir–Blodgett, layers can be found elsewhere [50,63]. TIRF spectroscopy was employed to selectively excite this single Trp residue in adsorbed hGH molecules and subsequently measure its fluorescence emission. The Trp residue is located in the hydrophobic interior of the hGH molecule and is H-bonded to another ␣-helix strand [64]. The location of the residue makes it a very sensitive local probe for the protein’s conformational changes. The effect of adsorption on the Trp fluorescence emission is demonstrated in Fig. 4a in which both the fluorescence spectrum of hGH in solution and adsorbed onto a methylated silica surface are shown. The fluorescence emission from the adsorbed hGH is blue shifted, indicating that either

FIG. 4 (a) Fluorescence spectra of hGH in solution (⫹) and after 0.5 h adsorption onto a methylated silica surface (x). hGH was dissolved in 0.15 M PBS at pH 7. (b) Stern–Volmer plot of quenching by trichloroethanol of hGH fluorescence in solution (squares) and adsorbed onto a methylated silica surface (circles). hGH was dissolved in 0.15 M PBS at pH 7. (From Ref. 50.)



451

the H-bonding between the Trp residue and the nearby ␣-helix has been disrupted or the process of dipolar relaxation around the residue has extracted less energy from the excited state (see Section II.A). One should note the importance of the position of this single Trp residue in hGH; similar conformational changes in other proteins would be more difficult to follow if their Trp residue were located on the protein exterior or if more than one Trp residues were present in the protein molecule. The use of fluorescence quenchers provides additional information about the conformation of adsorbed hGH. Recall that the quencher has to come in close contact with the fluorescence probe in order to absorb the energy of the excited state. In the case of hGH, the quencher has to penetrate into the protein structure to reach the single Trp residue. Hence, the changes in fluorescence quenching reveal the changes in the permeability of a particular hGH conformation. In the present example, quenching experiments are performed by using increasing concentrations of trichloroethanol dissolved in buffer solution. Collisional quenching of hGH fluorescence is described by the Stern–Volmer equation [18], F0 /F = 1 ⫹ K[Q], in which F0 and F are the fluorescence intensities in the absence and presence of the quencher, respectively, [Q] is the concentration of the quencher, and K is the Stern–Volmer constant. The hGH experimental quenching data are shown as a Stern–Volmer plot in Fig. 4b for the protein molecule in solution and adsorbed onto the methylated surface. The difference between slopes in Fig. 4b clearly indicates that the quenching of the adsorbed hGH is characterized by a larger Stern–Volmer constant; i.e., the quenching is more efficient at any given concentration of the quencher for adsorbed than for dissolved protein. The combination of the emission (Fig. 4a) and quenching (Fig. 4b) results indicates that the conformation around the tryptophan residue has been changed and the permeability of hGH conformation for trichloroethanol has been considerably increased upon adsorption on a hydrophobic surface.

III.

INFRARED SPECTROSCOPY AND CIRCULAR DICHROISM AS GLOBAL PROBES OF PROTEIN MOLECULES

Although local protein probe spectroscopy techniques such as fluorescence can be used to monitor protein conformational changes, the information obtained is restricted to the microenvironment of the fluorescent probe(s). In order to obtain more global information on the adsorbed protein secondary structure, the following two spectroscopic techniques are frequently used: infrared (IR) spectroscopy and circular dichroism (CD). Both techniques are based on the absorbance of light by the protein molecule. Different variants of infrared spectroscopy are nowadays used in surface adsorption studies. In external reflection–adsorption IR spectroscopy, for example, one uses the geometry of external reflection to measure the reflectance changes by the surface films. The internal reflection IR spectroscopy relies on various types of internal reflection elements to conduct light to the interface where the surface film attenuates its intensity by absorption.* *The field is rich in acronyms. By combining the words infrared, reflection, adsorption, external, spectroscopy, glancing, attenuated, total, and Fourier transform, one can explain all acronyms found in the literature: RAIS, IRAS, FT-IRAS, IRRS, IRRAS, RAIRS, FTIR, ATR, ATR-FT-IR, GIR, IR-ERS, RAS, RAIR, and perhaps a few others [5].


452

A.

Hlady and Buijs

Infrared Spectroscopy

In IR spectroscopy, one utilizes the vibrational states of chemical bonds, especially those in the protein backbone, to study the protein structure. The mid-infrared region between 700 and 4000 cm⫺1 is the spectral region most frequently studied. The single vibrational energy is usually accompanied by a number of rotational energy transitions which broaden the absorption band. In the case of macromolecules the contributions from many atoms to a particular vibration complicates the analysis of the IR spectrum. The analysis of the spectral region between 1100 and 1700 cm⫺1 has proven to be a valuable tool to gain information on global properties of the polypeptide conformation and is actually one of the earliest methods employed to study the secondary structure of proteins [65]. The amide IR bands of polypeptides consist of nine characteristic vibrational modes or group frequencies [66]. One of the most useful is the amide I band, located in the wavenumber region approximately between 1600 and 1700 cm⫺1 (i.e., at wavelengths around 6 ␮m). The amide I band primarily represents the C — —O stretching vibrations of this chemical group in the protein backbone (coupled to the in-plane N — H bending and C — N stretching modes). The frequency of this vibration depends on the nature of the hydrogen bonding in which the C — —O group participates. This, in turn, is highly sensitive to the secondary structure adopted by the polypeptide chain, e.g., ␣-helices, ␤-sheets, turns, and random coil structures and thus provides fingerprints for the protein secondary structure elements [16]. The characteristic frequencies of the different types of secondary structures have been established as a result of systematic studies on a number of protein samples [66–68] and by theoretical calculations [69–71]. In principle, it is also possible to study the protein structure using the amide II band, which represents the N — H bending, and the amide III band, which primarily represents absorption from chemical groups in the side chains. The sensitivity of the amide II region for the polypeptide conformation is relatively small. The amide II region can be useful, however, for quantifying the amount adsorbed [72–74]. The use of the amide III region (1100–1500 cm⫺1), although sensitive to the secondary nature, is troublesome because these bands are rather weak. Still, the amide II and III regions can be employed to obtain additional information on the results generated by analysis of the amide I region [75,76]. The amide I and II bands in an IR spectrum if IgG molecules adsorbed on a silicon cylindrical internal reflectance crystal are shown in Fig. 5a. Although IR spectroscopy is one of the earliest methods used for estimating the protein structure, its practical use was limited until the development of the interferometer-based, Fourier-transform infrared (FTIR) instrumentation. FTIR has two

> FIG. 5 (a) Infrared absorbance spectrum in the amide I and amide II spectral regions of IgG molecules adsorbed on a silica surface. The spectrum consists of an average of 300 scans and is smoothed by a Gaussian filtering over a frequency range of 35 cm⫺1 [118]. The IgG molecules are adsorbed for 1 h from an aqueous solution containing 50 ␮g/mL IgG and 5 mM phosphate buffer at pH 6. (b) Second-derivative spectrum of the amide I spectral region of adsorbed IgG molecules shown in (a). (c) Curve fitting of individual absorption bands (dashed lines) to the original amide I spectral region (solid line) shown in (a). The curve-fit procedure assumed a Lorentzian shape of the line broadening and is described in Ref. 72.




453

454

Hlady and Buijs

distinct advantages over the classical IR spectroscopy instruments which were based on the grating monochromators: (1) multiplexing advantage because of the simultaneous detection of all frequencies and (2) luminosity advantage because of the higher light throughput of interferometers over the grating monochromators. Consequently, the IR spectra are acquired more rapidly and with a better signal-to-noise ratio with FTIR instruments. The high sensitivity of FTIR in combination with the totally internally reflected IR beam, commonly referred to as attenuated total reflectance (ATR), offer a valuable method for studying proteins at interfaces [77,78]. The ATR-FTIR technique has proven its value in the analysis of secondary structure elements of proteins in close proximity of the sorbent surface. Examples include studies in which the secondary structure of adsorbed proteins were subjected to alterations due to different adsorbents [72,73,79,80], protein aggregation [81,82], various solvents [83,84], degree of solvation [85], and temperature variations [86]. Similar to the experiments performed with TIRF, ATR-FTIR is also used to access several different characteristics of the protein adsorption process. The infrared absorption intensity can be calibrated [87,88] and transformed in the adsorbed mass/ area amounts, which allows for the study of adsorption kinetics [72,79,89,90]. Since each protein has its specific structure and thus a specific infrared adsorption spectrum, an additional advantage of FTIR over TIRF is the possibility of studying adsorption from protein mixtures without any labeling requirements [91–93]. Furthermore, it is possible to use polarized excitation in an ATR-FTIR geometry and study the orientation and order of adsorbed proteins [94]. Even though ATR-FTIR spectroscopy has proven its value in the study of proteins at interfaces, it is important to realize that the technique also has some practical difficulties. For most research purposes, protein adsorption is preferentially studied in aqueous solutions and involves the problem of interference of the very strong and broad water absorption band around 1640 cm⫺1 with the amide I region. To overcome this problem many infrared studies on protein structures have been performed in D2O. Using modern sensitive instrumentation, however, it is possible to separate the intense water absorption band from the protein bands. Most of the infrared absorption from solution can be accounted for by subtracting the background spectrum of the solution from the protein absorption spectra. Still a more precise subtraction of solution signal is required as adsorbed proteins displace water molecules from the surface. Fluctuations in the physical properties of the system, such as temperature fluctuations, can alter the optical path length of the IR evanescent wave, thereby changing the intensity of absorption by solution molecules. The simplest and most often used method to overcome this problem is to use a scaling factor in the subtraction of the absorption by the solution utilizing a horizontal baseline in the spectral region where proteins do not absorb (1730–1800 cm⫺1). Somewhat more advanced algorithms, all based on the subtraction of the solution signal in the spectral region between 1720 and 2700 cm⫺1, have also been developed [95,96]. Other features which might complicate the analysis of the IR absorption bands sources are the traces of water vapor in the instrument’s cell compartment and the overlap between the amide I and amide II bands. The overlap can result in an overestimate of the amount of the structural component contribution commonly determined in the lower wavenumber region of the amide I band. In this same low wavenumber region it is quite likely that some of the IR energy is absorbed by side chains of the proteins [97]. The vibrational motions of the chemical groups can also be affected by the © 2003 by Marcel Dekker, Inc.


455

types of interactions these groups can have with the sorbent surface [98], although this is more likely to happen to the side chain groups than to the chemical groups in the protein backbone. B.

Analysis of Protein Conformation from IR Spectra

1. Resolution Enhancement Although the IR spectrum of an adsorbed protein contains all of the information on the protein’s secondary structure, the analysis of an FTIR spectrum for a qualitative estimation, and even moreso for a quantitative estimation, of the protein secondary structure is complicated because of the overlap in absorption bands originating from different structural components. The bands of a protein’s structural components are generally blended in one broad-shaped band in the experimentally measured spectrum, as can be seen in Fig. 5a. Improvement of the instrumental resolution is not helpful as the resolution is limited by the natural bandwidth of the absorption bands. However, mathematical procedures are available which will improve the discrimination between the individual bands. Two commonly used procedures for computational resolution enhancement are the second derivative and Fourier self-deconvolution calculations [99]. Both methods have successfully been applied in the analyses of the secondary structure of several proteins [67,68]. It should be emphasized that the resolution of the spectra is enhanced at the expense of a decreased signal-tonoise ratio, and great care has to be taken when analyzing the resolution-enhanced spectra. Besides mathematical resolution enhancement it is also possible to separate overlapping bands by adding experimental variables [75]. 2.

Peak Assignment to Structural Elements

After establishing the individual absorption bands, the next step is to relate those bands unambiguously to the secondary structure components. This is generally not a problem for a few major structural components, but it might be more troublesome for some smaller peaks located at both extremes of the amide I region. The amide I components associated with ␤-strands are located in the low wavenumber range between 1620 and 1640 cm⫺1, and a peak around 1635 cm⫺1 is generally ascribed to antiparallel ␤-sheets. Components in the lower region may originate from subtle differences in hydrogen bonding patterns in the antiparallel ␤-strands or from parallel ␤-sheets [66,67]. In addition, the side chain absorption could contribute to absorption in the low wavenumber part of the amide I region [97]. In most FTIR studies, resolution-enhanced protein spectra show a peak close to 1644 cm⫺1, which is generally assigned to random coil segments. A component around 1654 cm⫺1 is frequently observed and associated with ␣-helices. Absorption bands of turns are located in the high wavenumber domain [71], but components associated with ␤-sheets can also be found in this region. An exact distinction between the location of turns and ␤-sheets in this high wavenumber domain is not yet unambiguously established. Note that the above-mentioned peak assignments are based on proteins in an aqueous environment and that the frequency position of the absorption bands can shift in D2O or other solutions. A second-derivative spectrum of IgG adsorbed from aqueous solution onto a silica surface is shown in Fig. 5b. IgG is known to have a high content of ␤-sheet structure whose absorption peak is clearly visible as a minimum © 2003 by Marcel Dekker, Inc.

456

Hlady and Buijs

in the second-derivative spectrum. A somewhat different approach has been used in the case of dried protein films on metal surfaces which were studied by an external IR reflection technique (FT-IRAS) [100,101]. The peak assignments have been aided by an independent IRAS study of amino acid adsorbed on metal surfaces [102,103]. However, the inferred orientation of particular protein groups, such as carboxylate, could also have been created by the process of drying and preparing the film for external reflection spectroscopy. 3. Quantitative Analysis Resolution-enhanced infrared spectra allow a qualitative and semiquantitative identification of the various secondary structures present in the adsorbed proteins. A more quantitative analysis can be performed using curve-fitting methods. The general procedure is to decompose the spectrum into its constituent bands, specifying their shapes, positions, and amplitudes by use of some optimization technique. Then the experimental spectrum is expressed as a superposition of a number of individual absorption bands. For the vibrational band it is generally assumed that they have either a Gaussian or Lorentzian shape or that the spectrum is a convolution of the two. Each Gaussian (and/or Lorentzian) band is initially positioned at the frequency position of the band as observed in the second derivative or Fourier self-deconvolved IR spectrum, followed by an iterative band-fitting procedure. The fitted individual bands are then assigned to the different secondary structure elements and subsequently quantified by using the integrated areas of the individual peaks. The last step of the analysis implies that the extinction coefficients of the structural components are assumed to be equal. This assumption has been validated in an extensive study with 21 proteins in which a good correlation was found between the secondary structures as obtained from FTIR spectroscopy and x-ray crystallographic data, respectively [68]. Resolution-enhanced spectra are mainly used to determine the frequency position of the individual structural components. However, spectral deconvolution can also be used to estimate the bandwidth by varying the input parameters of the deconvolution process. Information on the bandwidth and absorption intensity of the individual bands can also be derived from the second-derivative spectra [72]. Bandwidths of the individual structural components are 15–28 cm⫺1. Second-derivative or deconvolved spectra of proteins often reveal that the amide I band consists of six or seven components [68,72,76,104]. Nevertheless, many interpretations are focused on two or three major components. Fitting an absorption band to as many individual bands as can be found might result in a more exact outcome of the quantitative amount of the various structural components. However, it also increases the chance of obtaining physically incorrect results in the fitting procedure. Fluctuations in the spectra which are not correlated to specific chemical structures are also enhanced in the deconvolved and second-derivative spectra and could accidentally be ascribed to an individual absorption band. Furthermore, one needs to take into account that more fitting parameters generally yield a better fit to the original spectrum but do not necessarily lead to a better physical picture of adsorbed components. As quantitative analysis only enhances the possibility of erroneous results, careful spectral interpretation is required. Hence, FTIR spectroscopy is more safely employed to follow protein conformational changes than to quantify protein structural elements in absolute amounts. © 2003 by Marcel Dekker, Inc.


457

In Fig. 5c, the result of a fitting procedure to the IR spectrum of adsorbed IgG (Fig. 5a) is shown [72]. For this fitting procedure, Lorentzian band shapes were assumed and the initial frequency position, bandwidth, and intensity were obtained from the second-derivative spectrum as shown in Fig. 5b. Assignment of the individual peaks to the protein’s structural components and quantification of the integrated areas of the individual absorption bands can then be used to study the secondary structure of adsorbed proteins. 4.

Adsorption-Induced Conformational Changes in IgG Studied by FTIR Spectroscopy

As an example of an FTIR study of global, secondary structure of adsorbed protein we cite the study performed on the monoclonal IgG molecules and their F(ab⬘)2 fragments whose adsorption was monitored as a function of solution pH [72]. The FTIR results were used to obtain the equilibrium amount adsorbed and the relative amount of random coil structure as concluded from the integrated area of the resolved peak at 1645 cm⫺1. The IgG molecules were adsorbed onto a circular silicon crystal from a flowing solution containing 50 ␮g/mL protein and 5 mM buffer. The equilibrium adsorbed amounts are shown in Fig. 6a, and the percentage of random coil structure is shown in Fig. 6b. A maximum in the amount adsorbed occurred near the isoelectric point (iep) of the proteins (which is at pH 5.8 for the monoclonal IgG and at pH 5.9 for its F(ab⬘)2 fragment). The maximum adsorption is often found around the protein’s iep, and it is generally explained by the effect of a low net charge density on the protein which results in a relatively high structural stability and hence a more compact protein structure in the adsorbed state. In addition, a decreased electrostatic repulsion between protein molecules allows for a tight protein packing arrangement on the surface. A higher protein net charge density is expected to result in a lower structural stability caused by internal electrostatic repulsion. This hypothesis is supported by the analysis of the adsorbed IgG IR spectra, which show that the amount of random coil structure increases when the pH is further away from the iep. Note that the net charge–related increase in the random coil structure is much larger for the whole IgG than for its F(ab⬘)2 fragment, indicating that the Fc part of the IgG molecule loses more of its structural conformation than the F(ab⬘)2 part. It is well known that under conditions of electrostatic repulsion between protein and a hydrophilic surface, the protein primarily adsorbs due to an entropy gain resulting from an increase in its random coil structure. Thus, the lower amounts of adsorbed F(ab⬘)2 fragments at the pH values above the iep might be related to the effect that the adsorbed F(ab⬘)2 molecules experience a smaller adsorption-induced increase in their random coil structure. C.

Circular Dichroism

As described in the introduction, spectroscopic techniques employed to study the adsorption and conformation of proteins at interfaces are often based on the concept of an electromagnetic field enhancement to selectively excite proteins in close proximity of surfaces. It is, however, also possible to obtain information on the adsorbed state of proteins using techniques and instrumentation originally developed to study proteins in solution, like circular dichroism or spectrofluorometry. Circular dichroism © 2003 by Marcel Dekker, Inc.

458

Hlady and Buijs

FIG. 6 (a) Adsorbed amount of (b) relative amount of random coil structure of IgG (open circles) and their F(ab⬘)2 fragments (closed circles) adsorbed at a silica surface at different pH values. The equilibrium adsorbed amounts were obtained after 6 h of adsorption, while the random coil structures were obtained for proteins which were adsorbed for an average time of 1 h. The isoelectric point (iep) is indicated on the x-axis. (From Ref. 72.)

(CD) is frequently used to address problems concerning global properties of the protein conformation and is especially sensitive to protein molecule ␣-helix content [105]. Because of the large number of successful CD studies on protein conformation in solution, the question arises how to most efficiently employ CD spectroscopy to study the conformation of adsorbed proteins. Indirect information on adsorptioninduced conformational changes has been obtained by applying CD spectroscopy to proteins desorbed from a surface [106,107]. To study proteins in situ in the adsorbed state a sorbent surface has to be present in the sample volume interrogated by the CD light beam. This sorbent surface should not interfere with the CD measurement (i.e., it should be transparent) and should adsorb enough proteins to obtain a reasonable signal-to-noise ratio. To meet these requirements, experiments have been initially performed using the geometry of stacked quartz plates [108]. © 2003 by Marcel Dekker, Inc.


459

Another way to record CD spectra from adsorbed proteins is to employ small colloidal particles, often referred to as ultrafine or nanoparticles, which, when suspended in solution, can serve as a sorbent material for proteins. Because of their small size (760–460 and 14–4.5 kD, respectively, possessed a high surface activity. These fractions are expected to contain high-molecular-weight glycoproteins and proline-rich proteins, statherin, cystatins, and histatins, respectively (see above). Upon adsorption to hydrophobic surfaces both fractions showed extensive adsorption and the lower-molecular-weight fraction was also found to dominate the adsorption at air–liquid interfaces as deduced from surface tension measurements. On pure silica, on the other hand, the highest adsorption was observed from the high-molecular-weight fraction. In order to obtain affinities for HA, adsorption of many proteinaceous salivary components has been studied. This has been done for amino acids [132], polymeric amino acids [150], salivary proteins [151], and glycoproteins [152]. More detailed studies on the salivary phosphoproteins have been carried out by Moreno, Hay, and coworkers [133,134]. In the first of these studies, adsorption of statherin and PRP-3 to HA, fluoroapatite (FA), and fluorohydroxyapatites was investigated by a solution depletion technique. It was found that the adsorption affinity for these materials was considerably higher for PRP-3 than for statherin. On the other hand higher numbers of statherin molecules could be bound as expected due its smaller size. Furthermore, they found that the adsorption affinity increased with the fluoride content in the mineral, which was attributed to a decrease in the surface energy of the adsorbent. In a following study the adsorption of PRP-1, PRP-2, PRP-3, and PRP-4 as well as the 30 residue N-terminal portion of these proteins [134] was investigated. The ad© 2003 by Marcel Dekker, Inc.

Protein Adsorption in the Oral Environment

831

FIG. 12 (a) The normalized force versus distance between mica surfaces 20 h after the addition of 17 ␮g/mL (1.05 ⫻ 10⫺6 M) PRP-1 in 1 mM NaCl (●) and in 1 mM CaCl2 (䊱) as well as the curve recorded 20 h after the subsequent addition of 10% saliva to a layer of PRP-1 preadsorbed from 1 mM CaCl2 solution (䡲) and in pure 10% saliva (䊲). The figure shows the force curves on a linear scale, and curves on compression and decompression are indicated by filled and unfilled symbols, respectively. (From Ref. 154.) (b) The force curves on compression are shown on a logarithmic scale. A linear fit was performed on the longrange part of the force and the resulting lines are inserted. An electrostatic repulsive force is roughly proportional to e⫺␬D, where ␬⫺1 is the Debye screening length. The resulting slopes ˚ (lower dashed line) for for the linear fits correspond to the following values of ␬⫺1: 86 A ˚ (lower full line) for PRP-1 in 1 mM CaCl2, 129 A ˚ (upper full PRP-1 in 1 mM NaCl, 72 A ˚ (upper dashed line) for pure line) for saliva added after PRP-1 preadsorption, and 143 A saliva solution. (From Ref. 154.)


832

Arnebrant

sorption was found to be endothermic and hence driven by increase in entropy. PRP1 and PRP-3 along with the N-terminal peptide showed higher affinity for HA than did PRP-2 and PRP-4, indicating an influence by the residue at position 50 being aspargine in the case of PRP-1 and -3 and aspartic acid in PRP-2 and -4. An unexpected finding was that a lower number of short PRPs were accommodated on the surface at saturation. As for statherin, mentioned above, the adsorption affinity was found to be higher for FA than HA, again ascribed to differences in surface energies. Their data showed that adsorption of PRP-1 and PRP-4 to HA reached plateaus at concentrations above approximately 15 ␮g/mL. Assuming a content of 30% PRP in HWS this means that these proteins will be able to saturate the surface at a saliva dilution of 40 times. Their work on salivary protein adsorption onto Ca apatites is summarized in Ref. 153. In a paper by Johnsson [25] data for adsorption to HA of salivary proteins are compiled (Fig. 13). Binding data were evaluated according to a Langmuir model, and binding constants are reported. They concluded that the highest affinities for HA were observed for statherin and the PRPs. Interestingly, the N-terminal portion of PRP-1 (PRP1T1) was found to show higher binding affinity than intact PRPs, whereas the dephosphorylated peptide showed a binding constant about 10% of the phosphoserinecontaining one. In line with this, it is suggested that the high-affinity binding to the HA surface takes place through interactions between negatively charged peptide sequences and positive surface sites. Figure 13 illustrates the differences in affinity and amounts and shows that adsorption appears to reach plateaus at concentrations below 50 ␮g/mL for all proteins investigated and that the amounts are in the range approximately corresponding to monolayers. In the same paper the adsorption is correlated to effects on HA crystal growth, where PRPs and cystatins were shown to be most efficient. In a recent study the nucleation of Ca phosphate at a HA surface was monitored in situ and found to be

FIG. 13 Adsorption to hydroxyapatite versus concentration of various salivary proteins. (Compiled in Ref. 25; PRP-1 from Ref. 134; amylase from Ref. 244; statherin and cystatin S and SN from Ref. 245; HSA from Ref. 246; and histatin 5 from Richardson, unpublished.)



833

slowed down in the presence of a salivary pellicle [82]. In the same study it was also shown that the inhibitory effect increased with pellicle age (Fig. 14). Adsorption of PRP-1, PRP-3, and statherin onto hydrophilic silica and hydrophobic model surfaces was recently studied in situ by ellipsometry [122]. In this study adsorption was found to be controlled by mass transport at concentrations up to the highest investigated (17 ␮g/mL). Furthermore, it was found that, except for PRP-1, much lower amounts were adsorbed on negatively charged hydrophilic surfaces as compared to hydrophobic ones. The concentration dependence of adsorption agrees well with previously observed behavior on HA, but maximum amounts adsorbed were 0.5–0.7 mg/m2, which is less than half a monolayer side-on (assuming the area of a truncated ␤-casein with molecular weights corresponding to PRPs). Crucial importance of interactions involving the binding of phosphoserines to surface Ca2⫹ may be one reason for the differences with respect to the behavior on HA. In addition, uncertainty may exist in the determination of the accessible surface area for HA particles. In a study of PRP-1 adsorption to mica and measurement of the interactions between adsorbed layers (see Figs. 12a and b) only weak adsorption was observed in 1 mM NaCl, and it was demonstrated that formation of a stable adsorbed layer relied on Ca2⫹ [154]. In the presence of Ca2⫹ the data indicated a biphasic structure with a dense inner structure and an outer compressible one. The thickness of the ˚ and could be compressed to approxfully developed layer was approximately 60 A ˚ imately 40 A at high load. It may be tempting to assign the inner and outer fractions of the adsorbed layer to the N- and C-terminal portions of the protein as in the orientation previously suggested for adsorbed layers on apatitic surfaces. The adsorption and interaction behaviors resemble those of ␤-casein which is structurally related and has been characterized by surface force measurements [155] and by ellipsometric and neutron reflectivity measurements [156].

FIG. 14 The figure shows precipitation/crystallization from 10 mM sodium phosphate buffer containing 1 mM CaCl2. Adsorbed amount versus time upon simultaneous addition of 10% saliva and 1 mM Ca2⫹ (open diamonds), upon addition of 1 mM Ca2⫹ after formation of a 1 h pellicle (filled diamonds), upon addition of 1 mM Ca2⫹ after formation of a 1 h pellicle and rinsing (open squares), and upon addition of 1 mM Ca2⫹ after formation of a 1 h pellicle subjected to rinsing and then aged for 12 h (filled squares). (From Ref. 82.)


834

Arnebrant

Due to its importance for protection [56,157], role in bacterial adhesion [6,158], and also relevance for mucoadhesion [159] adsorption of mucins has been studied with respect to a variety of surfaces and by applying different techniques [39,108,148,149,159–162]. Lindh et al. investigated the adsorption of purified MUC5B mucin (MG-1) to hydrophilic and hydrophobic surfaces and compared its adsorption behavior with commercial bovine submaxillary mucin (BSM) [108]. The adsorption affinity of both mucins was found to be highest for hydrophobic surfaces. Adsorbed amounts of pure MUC5B were found to be 1.3 and 2.7 mg/m2 at hydrophilic and hydrophobic surfaces, respectively, at a solution concentration of 0.1 mg/ mL. Furthermore, adsorption of the MUC5B and the commercial preparation was found to agree well at low concentration, whereas at higher concentrations the commercial sample adsorbed less to the hydrophobic and more to the hydrophilic surface compared to the MUC5B, which may be ascribed to possible presence of proteinaceous impurities. It was also found that adsorption kinetics were slow and that adsorption was mass transport limited at low concentrations. Tabak et al. [39] investigated the adsorption of MG-1 and MG-2 to HA surfaces and evaluated the data in terms of a Langmuir model. A higher affinity for HA by MG-1 was reported, and furthermore it was found that the adsorption of MG-1 was inhibited by salivary glycolipids but was unaffected by a cysteine-containing salivary glycoprotein. Surface force measurements on BSM adsorbed on mica [148] and gastric mucins on hydrophobized mica [149] reveal that long-range steric repulsion dominate the interactions between such layers. This observation is relevant for the barrier function of mucous layers, and the resemblance to the interaction between salivary films mentioned above indicates that this function may be successfully fulfilled at oral interfaces. C.

Aspects of Composition and Development

It was early shown by histological staining that the pellicle was of proteinaceous nature. Furthermore, these observations were supported by the finding that the pellicle was removed upon exposure to proteolytic enzymes [163]. This was much later demonstrated in a real-time in situ experiment showing the degradation of an in vitro pellicle on silica by a mixture of proteolytic enzymes [123]. In an early report by Armstrong the composition of the pellicle was reported to be 46% amino acids, about 3% hexosamines, and a total of 14% carbohydrates [164]. The knowledge on the content and composition of lipids is less comprehensive, but cholesterol; cholesteryl esters; tri-, di-, and monoglycerides; phospholipids such as phosphatidylcholine, phosphatidyletanolamine, and sphingomyelin along with glycolipids are reported to be present [165–167]. Even though the lipid content in saliva is low (see above) the pellicle may contain substantial amounts and at 2 h pellicles are reported to contain 22–23% lipids, mainly glycolipids but also neutral lipids and phospholipids [165,167]. Carbohydrates found are glucose, glucosamine, galactose, and mannose [168]. A high content of glucose has been observed, while the proteins bound to the HA contain mainly galactose, mannose, and fucose, but only minor amounts of glucose, indicating that its presence in pellicle does not originate from pellicle proteins [130]. In an early investigation of adsorption from HWS to HA and enamel powders in vitro as well as on freshly extracted teeth, Hay [12] demonstrated that the fraction © 2003 by Marcel Dekker, Inc.


835

of the adsorbed salivary proteins that was eluted in 0.2 M phosphate buffer showed a selective binding in both cases and displayed similar electrophoretic mobilities. Furthermore, only minor amounts of sialoproteins were detected on the extracted teeth. In later papers the same author showed that proline-rich proteins, sIgA, amylase, and albumin could be eluted from HA powders exposed to HPS by ethylene diamine tetraacetate (EDTA) and that the highest affinity components were acidic peptides with a high content in glutamic acid, tyrosine, aspartate, and histidine, respectively [130]. By experiments involving varying HA-to-saliva ratios, Hay further showed that the initial adsorption to HA was dominated by a tyrosine-rich peptide (statherin) [8] followed by histidine-rich peptides (histatins) [131], and then by proline-rich proteins [132]. Embery demonstrated that low-molecular-weight phosphoproteins played an important part in the initial in vivo pellicle formation but gradually disappeared with pellicle maturation [40]. Work on homo- and heteropolyaminoacid adsorption to HA has demonstrated the importance of phosphate ester groups [169] and dicarboxylic amino acids [150] for high-affinity binding. Studies of the initial adsorption kinetics upon in vitro pellicle formation from glandular saliva to model surfaces also support the finding of initial binding of low-molecular-weight peptides/proteins of the size corresponding to statherin and PRPs to model surfaces (see Fig. 6b above) [106,122]. As indicated above mucins are believed to be incorporated much later in the pellicle, as shown by, e.g., Busscher et al. [107], who demonstrated the selective adsorption of mucins from HWS in vitro and showed that they were absent within the first hour but appeared after 2 h and continued to increase in concentration during the next 16 h. From experimental observations in vitro and simple mass transport analysis it was also shown that mucins were not involved in the initial adsorption from HSMSLS on hydrophobic and hydrophilic model surfaces [108]. The amino acid composition in acquired pellicles has been analyzed in a number of studies. One of the first was carried out by Sonju and Rolla in 1973 [41]. Data from many authors were compiled by Lendenmann [17] and show good agreement between different reports irrespective of the pellicle age, location of sampling area, presence of plaque, etc. Amino acid compositions alone are, however, of limited value in the prediction of the protein composition. Among the observations are preferential accumulation of hydrophobic amino acids in pellicle as demonstrated by AlHashimi et al. [170] and a lower content of proline than in the whole secretion [17,18]. By elution by sodium dodecyl sulfate (SDS) soaked foam sponges and subsequent electrophoresis and immunoblotting techniques Carleń et al. have investigated pellicles formed in vitro on dental materials [147] and in vivo and in vitro from HWS, HPS, and HSMSLS [146]. From the latter study it was concluded that pellicle compositions were reflecting the composition of saliva prevailing in the part of the mouth where the pellicles were formed, an observation that may be pertinent to the subsequent adhesion of bacteria. Furthermore, albumin was observed clearly in vivo but not in vitro. By using a combination of amino acid analysis, electrophoresis, gel filtration, and HPLC, Jensen et al. studied the protein composition on pellicles formed in vitro on HA [9] from HWS, HPS, and HSMSLS. Amylase, acidic and glycosylated proline-rich proteins, statherins, and histatins were found in the HPS pellicle, and by the use of cationic electrophoresis histatin 3 and 5 were identified. The HSMSLS pellicle was identical except for the presence of cystatins and © 2003 by Marcel Dekker, Inc.

836

Arnebrant

absence of glycosylated PRPs. The HWS pellicle, on the other hand, was shown to contain mainly amylase, acidic PRPs, cystatins, and proteolytically derived peptides. In a study by Lamkin et al. the proteolytic degradation in HWS was further investigated and it was concluded that degradation was an ordered and consistent process and that the proteolytically derived fragments amounted to approximately 5% of the protein adsorbed to HA [171]. Yao and coworkers applied amino acid analysis, MALDI-TOF, and LC-MS to compare in vivo pellicles eluted by a swabbing procedure and in vitro HA pellicles from HWS [18]. They found that ammonium bicarbonate along with SDS were most efficient in removing pellicle components, removing 70 and 80%, respectively. In the in vitro pellicles lactoferrin, albumin, amylase A and B, PRP (Db, long, and short), lysozyme, cystatin SN, statherin, and peroxidase were identified. In vivo, on the other hand, the PRPs were absent and in addition low-molecular-weight peptides were observed in accordance with previous studies. It was argued that the differences were due to the different surface properties of enamel and HA and the presence of proteolytic enzymes in the in vivo case. By use of SEM, TEM, and immunological techniques the presence of PRPs, histatins, and statherin were shown to be integral parts of the in vitro pellicle on HA and bovine enamel [35]. TEM was also used by Busscher and coworkers in a study of buildup of in vitro pellicle from reconstituted lyophilized HWS on enamel in a flow cell system [172]. They observed that the characteristics of the enamel surface disappeared after seconds and that within 10 s of salivary exposure three or four distinct homogeneous films were deposited on top of each other and an uneven knotted structure developed. This heterogeneous pattern was observed for at least 2 h. TEM was also used by Hannig to investigate the ultrastructure of salivary pellicles formed after 6 h on different dental materials in vivo. He found that pellicles on test pieces exposed on the buccal side ranged from 500–1000 nm and displayed a globular and heterogeneous structure, whereas on the lingual side a granular pellicle with a thickness of about 100 nm was formed [173]. In a subsequent study it was shown that the in vivo pellicle was initiated by a quick formation of a dense layer followed by attachment of an outer loosely arranged layer. At 24 h the pellicle was also shown to be influenced by the oral environment rather than properties of different dental materials [97]. In view of the significance of the function of each of the pellicle proteins, for example, the reported function of the precursor proteins in mineralization, lubrication, and bacterial modulation, their presence at the enamel surface may be of vital importance at specific and different time points. The heterogeneous protein composition of saliva, including proteins at varying concentrations with a wide range of molecular weights and structures and hence diffusion rates, means that exchange phenomena should be anticipated. Dynamic adsorption behavior featuring exchange reactions are known to occur in the case of synthetic polymer mixtures or polydisperse samples (Chapter 1). For proteins they have been demonstrated in the field of plasma protein interactions with blood contact surfaces [13,16,174,175]. In the latter system, the sequences involving a progression from low-molecular-weight proteins at high concentration followed by proteins with a higher affinity and molecular weight have been demonstrated. In saliva analogous behavior may take place, which is supported by the reported affinity constants for HA binding [25] which indicate that, e.g., mucins should have a higher affinity for HA than phosphoproteins and thereby have the potential to © 2003 by Marcel Dekker, Inc.


837

replace the initially bound statherin and PRPs. The fact that this takes place in reality is supported by the experimental observations reviewed above. It is hence clear that at least in vivo the pellicle develops from essentially a monolayer of adsorbed salivary proteins formed within seconds to much thicker structures within the time frame of hours. Based on the compositional data presented above it may be reasonable to assume that the sequence in the case of saliva will initially involve the lowmolecular-weight phosphoproteins and after this globular proteins adsorb as do large glycoproteins (agglutinins, mucins), thereby forming a complex mixed layer. It may be speculated that the content of large glycoproteins increases with time during this phase. Finally, after long times in contact with saliva, aggregates from solution such as the micelles described by Rykke et al. may attach. A mechanism along similar lines has previously been proposed by Embery [176].

V.

PLAQUE FORMATION

One definition of dental plaque is ‘‘the nonmineralized microbial accumulation that adheres tenaciously to tooth surface, restorations, and prosthetic appliances, shows structural organization with predominance of filamentous forms, is composed of an organic matrix derived from salivary glycoproteins and extracellular microbial products, and cannot be removed by rinsing or water spray’’ [177]. Dental plaque also fulfils the criterion of being a typical biofilm [178,179]. Plaque is associated with the major dental disorders caries and periodontal diseases. A caries lesion is the result of demineralization of dental hard tissue caused by organic acids produced by fermentation of carbohydrates like sucrose, glucose, and fructose by oral bacteria. For enamel it occurs below a critical pH of about 5.5 and starts a few minutes after carbohydrate consumption and lasts for approximately 30 min. For dentin the critical pH is as high as pH 6.7. The presence of plaque imposes diffusion restrictions resulting in local pH drops and increased duration. Caries is associated with bacteria such as Streptococci, Lactobacilli, and Actinomyces strains. Periodontal diseases include gingivitis and periodontitis, both caused by accumulation of plaque. Gingivitis is a reversible state characterized by a swollen gingiva, increase in pocket depth, and bleeding, whereas periodontitis also involves alveolar bone resorption and detachment of periodontal ligaments, which results in irreversible tissue damage. Associated bacteria are gram negative and mainly anaerobic, e.g., Porphyromonas, Bacteroides, Fusobactrium, Capnocytophaga, and Spirochetes, which produce extracellular enzymes and toxins that destroy connective tissue and contribute to the inflammatory response. For up-to-date reviews on the structure, composition, and buildup mechanisms of plaque biofilms, see Refs. 178, 180, and 181. The initiation of plaque formation takes place with the adhesion of microorganisms, predominantly bacteria, onto oral interfaces such as the surfaces of teeth or restorative materials. This occurs rapidly and bacteria have been observed at germanium prisms exposed intraorally after 2 h [95] and are reported to be present at enamel surfaces after 4 h of in vivo exposure [182]. The primary colonizers are Streptococcus, Actinomyces, and Veillonella strains. © 2003 by Marcel Dekker, Inc.

838

Arnebrant

Adhesion may take place to clean or pellicle-coated surfaces, and due to the fast rate of protein deposition the latter is the normal. Protein adsorption will monitor adhesion of plaque bacteria through specific and nonspecific means (see below), and it has been shown that the initial adhesion strength depends on whether adhesion takes place on a clean surface or via a protein film [88]. Plaque composition is known to be different at different loci, for example, in supra and subgingival plaque, and furthermore a sequential pattern in the colonization has been reported to be present. Mechanisms behind the adhesion may be of specific and nonspecific nature, where the latter may be approached from a colloid chemical perspective [183,184]. The specific route involves adhesin–receptor (ligand on oral surfaces) interactions, possibly including conformational changes, and high-affinity binding. For reviews see Refs. 181 and 185. Both will have to be considered, as illustrated in Fig. 15, but their relative importance may vary [186]. The nonspecific adhesion stemming from general considerations of electric double layer and van der Waals forces (DLVO theory) may be predicted from data on available charge densities, sizes, and properties of the surrounding medium [183]. Along this route methods have been developed to quantify surface properties of oral interfaces [128,187,188] and bacteria [189–191]. The role of surface energies in determining adhesion strengths in this context was early recognized by Zisman [192], and later Baier, Shafrin, and Zisman [193] discussed the relevance of the critical surface tension in order to predict outcomes of biological adhesion events. This concept was employed by Glantz, who showed that the maximum amount of adhered plaque to tooth surfaces was dependent on the surface free energy of the solid as quantified by contact angle measurements [86]. Baier and Glantz [128,143,187,194] later showed that a range in critical surface tension between 20–30 mN/m exhibited low biological adhesion and was hence denoted ‘‘bioabhesive.’’ Regarding specific interactions it has been demonstrated that the adhesion of oral bacteria is mediated by interactions between adhesins on the bacterium surface or on fimbriae [195] and salivary components such as mucins [158], salivary agglu-

FIG. 15 Nonspecific (left) and specific (right) interactions upon bacterial adhesion. The former may be van der Waals and electrostatic double layer forces originating from the entire interacting surfaces. The latter, specific interactions may contain electrostatic, van der Waals, and hydrogen bond components, but take place between highly localized chemical groups. (From Ref. 247.)



839

tinins [176,196], statherin, and proline-rich proteins [197,198]. The affinity for the latter by Streptococcus and Actinomyces strains was early demonstrated by Gibbons [197–200]. An interesting feature of microorganisms is their ability to produce biosurfactants as defense against other colonizing species. This was demonstrated experimentally for Streptococcus mitis, which produces a biosurfactant that is able to substantially affect the adhesion of Streptococcus mutans [201]. The surfactant was found to be a glycolipid that was capable of reducing surface tension of aqueous solutions to 30–40 mN/m. Considering the presence of polar lipids in the pellicle, including reported interactions between MG-1 and glycolipids [39], the role of surfactants and polar lipids in oral adhesive events should not be neglected. As outlined above the dental plaque may be considered a biofilm in the sense that it is a dynamic vital ecological unit that comprises a variety of bacterial accumulations that grow on an inert or biologically active surface. The plaque is subject to development and adaptation to changes in environmental conditions through changes in the relation between the more than 500 species found in a plaque [181]. It is clear that after the initial adhesion of the primary colonizers, which takes place in a sequential fashion, these species multiply and when the available surface is filled growth will continue normal to the surface in ‘‘towers’’ (see Ref. 180 and references therein). After approximately the third day filamentous bacteria are reported to be present in the predominantly coccoid plaque and corncob structures are detectable [180]. Figure 16 shows a hypothetical picture of the initial stages in bacterial adhesion and how these may affect the composition and structure of the biofilm. After about a week filamentous bacteria penetrate the coccoid plaque and starts to replace it, and after about 2 weeks columnar microbial colonies are replaced by a dense mat of filamentous colonies. In the process of coadhesion of bacteria the same reasoning

FIG. 16 Hypothetical picture illustrating initial bacterial adhesion and its role in determining biofilm growth. (Adapted from Ref. 247.)


840

Arnebrant

as above would apply involving nonspecific and specific mechanisms. As in initial adhesion, adhesin-receptor mechanisms have been suggested to be involved [181]. Out of the currently identified models for biofilm structure—that is, water channel, heterogeneous mosaic, and dense biofilm model—the latter was suggested to conform with oral biofilms by Wimpenny and Colosanti [202]. Important components in biofilms in general, including plaque, are extracellular polysaccharides (EPS), which are important for the integrity of the biofilm and may comprise 50–95% of their dry weight [203]. The EPS produced differs between species and by accessibility of nutrient in oral biofilms; for example, glucans, fructans, and levans are common. An interesting feature of bacteria in biofilms is their ability to adapt to the surroundings in the sense that a different phenotype may be expressed [179]. One consequence of this is their tolerance to antibiotics and antimicrobial substances, which may be a factor of 1000 higher [204]. Furthermore, the topic of intercell signalling by low-molecular-weight substances like homoserine lactone within communities of microorganisms, as demonstrated by Davies and coworkers [205,206], has received considerable attention. Furthermore, it has been shown that protein expression is different in planktonic and biofilm states. For recent reviews on oral biofilms, see, e.g., Refs. 178, 207, and 208. In the literature a multitude of model systems have been developed and presented for studies of initial biofilm buildup [126,127,129], microbial adhesion [68,209], controlled biofilm growth [210–213], and plaque growth [214]. An important aspect in this respect is the possibility to study adhesion processes in situ without the complication of passing air–liquid interfaces.

VI.

ROUTES FOR MODULATION OF PLAQUE GROWTH

Reduction in plaque and measures to minimize consequences of the presence of dental plaque are of course of prime importance in the dental field. Several strategies have been identified to combat plaque-related diseases (for reviews, see, e.g., Refs. 215–217). Some possible routes are summarized below: 1. 2. 3. 4. 5.

Increasing the resistance of the teeth to decay Modifying the surface of the teeth in order to reduce bacterial adhesion Interfering with production of extracellular material Using bactericidal compounds Using antibiotics

Route 1 mainly involves treatment with fluorides, which from a surface chemical perspective also is known to affect the surface energy and hence adhesion when incorporated in enamel [86]. Routes 2–4, on the other hand, involve some interesting and challenging connections to the interfacial behavior of proteins. As in the field of blood contact surfaces effects of surface wettability have been addressed, and it has been shown experimentally that hydrophobic surfaces perform well with respect to accumulation of dental plaque in vivo [86,144]. Rolla and coworkers reported on the use of polydimethylsiloxane to reduce protein adsorption in vitro and in vivo [218]. Olsson et al. reported on a permanent hydrophobization by covalently attached silicone polymer, which was shown to reduce adsorption of model proteins and the © 2003 by Marcel Dekker, Inc.


841

amounts of stainable pellicle and plaque accumulated in vivo compared to untreated controls [219]. The potential of adhesion-reducing hydrophilizing compounds, of low molecular weight as well as polymeric nature, have been explored extensively over the last decades. Effects by surfactants in inhibiting bacterial adhesion have been investigated in a number of papers. Rolla and coworkers reported on the use of SDS [220], pyrophosphate [221], and phosphonates [222]. The effects of polyalkylene oxide derivatives on the binding of Streptococcus mutans in vitro was investigated by Olsson and coworkers, and it was found that both phosphates and phosphonates were able to reduce binding to pure HA but that the effect was limited on HA exposed to saliva [223]. Furthermore, efficient reduction in adhesion of Streptococcus mutans was accomplished in vitro by binary mixtures of surfactants involving alkyl phosphates aimed at binding to the HA surface and nonionic ethoxylates [81]. Coating of surfaces with hydrophilic noncharged polymers is an efficient route for reduction of protein adsorption provided that the layers are densely packed (Chapter 27). This concept was explored by Olsson and coworkers and compared to the previous approaches [135,224]. The potential of ethyl hydroxyethyl cellulose (EHEC) in unmodified and phosphated form was investigated, and the polymers were found to be effective adhesion reducers in absence of saliva. After preexposure of HA to saliva, however, effects were strongly reduced and neither approach gave significant plaque reductions in a small clinical trial [225]. Polyethylene imine (PEI)–polyethylene oxide (PEG) copolymers are often used to avoid nonspecific protein adsorption (Chapter 27). These polymers were shown to reduce protein adsorption and pellicle buildup almost completely, but showed much higher adhesion of Streptococcus mutans and in vivo plaque than hydrophobic reference surfaces prepared by plasma polymerization of hexamethyldisiloxane [135]. The surface modification work by Olsson, Holmberg, and coworkers is summarized in Ref. 226. Other means of modulation include binding of proteins in order to passivate the surfaces or direct subsequent binding of bacteria to harmless flora. One alternative that has been explored is the use of the 30 residue anchoring peptide of PRP-1 as a high-affinity coating for HA. Furthermore beneficial properties with respect to caries protection have been reported from milk and milk components. Early studies focused on the casein phosphopeptides such as the caseinoglycomacropeptide [227]. It was later shown that proteose-peptone fractions 3 and 5, which are proteolytically derived peptides from ␤-casein, were found to have a high binding affinity for dental enamel and possess a strong reduction in the demineralization rate [228]. This finding is interesting to relate to the proteolytical degradation of proline-rich proteins and presence of derived peptides in HWS. Johansson has found that statherin-specific antibodies were able to block statherin-induced binding of Candida albicans to HA and to reduce it on buccal epithelial cells [229]. Furthermore, Edgerton showed that histatin 5 adsorption to modified PMMA [230] increased its anticandidal activity. Concepts found in route 3 include interference with or modification of the exopolymer producing enzyme glycosyltransferase (GTF). This may be accomplished through inhibitors, substate analogs, and antibodies [215,231,232]. Another concept with interesting features is the antiplaque compound delmopinol, a surfactant having a low antimicrobial profile which is reported to influence several items mentioned in routes 2 and 3. For example, it has been shown to influence in vitro pellicles [118,233], wettability of tooth surfaces [233], surface potential © 2003 by Marcel Dekker, Inc.

842

Arnebrant

and flocculation of oral bacteria [234], production of extracellular polysaccharides [214,235], as well as in vitro plaque cohesion [214]. Route 4 includes use of antibacterial compounds such as chlorhexidine (CHX), cetylpyridinium chloride (CPC), and triclosan, which have been widely used in order to control oral biofilms and many studies have been carried out showing their effects on plaque formation and related diseases (cf. [178]). CHX and CPC have been formulated in a two-phase formulation based on 15% vegetable and essential oils and an aqueous phase containing CPC or CPC plus CHX, and the concept was shown to have a high capacity in desorbing adhered Acinetobacter calcoaceticus RAG-1 from polystyrene in vitro [236]. CPC and CHX are cationic surface active compounds which means that effects on pellicles may be anticipated and have been demonstrated for CHX in vitro [237]. Triclosan, which is an oil-soluble bisphenol, has been solubilized in micelles of Tween 80 and sodium dodecyl sulfate, and it has been demonstrated that only the SDS formulation showed significant clinical effects [238,239]. This finding might be related to the efficiency of pellicle displacement reported for this surfactant [82,119]. Furthermore, triclosan was also used in combination with silicone oil [240] with positive results on bacterial adhesion and plaque reduction in vivo. In routes 4 and 5, an attractive feature in antiplaque treatment is to increase the substantivity of the respective agent in the oral cavity, and therefore formulations of antibacterial compounds and antibiotics exhibiting sustained release properties are currently under development. Devices include various varnishes, strips, and gels to be inserted or applied on teeth or in gingival pockets (for a review, see Ref. 217). To mention an example, one interesting approach is a lipid-based formulation of metronidazole based on monoolein and triglyceride oil, where a dilution-induced phase transition from L2 (reversed micellar) to H2 (reversed hexagonal) is utilized in order to obtain in situ gelling properties.

VII.

SALIVARY SUBSTITUTES

An important area in the dental field is the development of saliva substitutes. The use of salivary substitutes is necessary for reduced salivary flow rates encountered as a consequence of systemic diseases like Sjo¨grens’s syndrome or caused by treatment of tumors by radio- or chemotherapy. Furthermore, many drugs have such negative side effects. Traditionally, highly viscous polysaccharides like carboxymethylcellulose (CMC) have been used in aqueous solutions in this application [93]. However, an increasing interest is directed toward the use of molecules originating in or resembling those performing the lubricating roles in saliva. Therefore potential types of molecules are high-molecular-weight glycoproteins, like mucins, which are responsible for the viscoelastic properties of saliva [57]. Rheological characterization of potential components in saliva substitutes such as polysaccharides [93], submaxillary mucins [241], mucin and albumin alone and mixed [92]. Lubrication has also been measured using a ‘‘tooth on disc’’ setup, and the lubricating properties for saliva and selected salivary substitutes were determined [89,91]. It was found that little correlation existed between viscosity and lubricating efficiency. It was also observed that a CMC-based product was found not to be effective in lubricating oral hard inter© 2003 by Marcel Dekker, Inc.


843

faces, whereas a mucin-based product was a better lubricant, resembling the properties of human saliva [89]. Thin film (boundary) lubrication was reported for commercial salivary substitutes containing CMC and mucin, respectively, whereas a glycerol-based product showed hydrodynamic lubrication [91]. Furthermore, on a molar basis lubricating efficiency of salivary molecules was found to decrease in the order MG1 > MG2 > nonglycosylated ␣-amylases ⬇ glycosylated ␣-amylases. In a paper by Reeh et al., enamel–enamel lubrication was assessed and lubricating properties of a number of CMC-, glycerol-, and mucin-based salivary substitutes reported [242]. In a recent paper comparing properties of commercially available saliva substitutes and saliva, the importance of adsorption and properties of adsorbed films for lubrication and duration (substantivity) were discussed [90]. Future development in this area may be expected to involve products replacing not only the lubricating properties of saliva, but also for example the ability to combat bacteria-mediated diseases [243].

ACKNOWLEDGMENTS The Institute for Research and Competence Holding (IRECO) and the Foundation for Knowledge and Competence Development (KK-stiftelsen) are acknowledged for financial support.

REFERENCES 1. 2. 3. 4.

5. 6.

7.

8. 9.

D. B. Ferguson, The salivary glands and their secretions. In Oral Bioscience (D. B. Ferguson, ed.). New York: Churchill Livingstone, 1999, pp. 117–150. B. L. Slomiany, V. L. N. Murty, and A. Slomiany, Salivary lipids in health and disease. Prog. Lipid Res. 24:311–324, 1985. G. N. Jenkins, The Physiology and Biochemistry of the Mouth, 4th ed. Oxford: Blackwell Scientific, 1978. D. I. Hay, Human glandular salivary proteins. In CRC Handbook of Experimental Aspects of Oral Biochemistry (E. P. Lazzari, ed.). Boca Raton: CRC Press, 1983, pp. 319–355. C. Dawes, G. N. Jenkins, and C. H. Tonge, The nomenclature of the integuments of the enamel surface of teeth. Br. Dent. J. 115:65–68, 1963. M. J. Levine, P. C. Jones, R. E. Loomis, M. S. Reddy, I. Al-Hashimi, E. J. Bergey, S. D. Bradway, R. E. Cohen, and L. A. Tabak, Functions of human saliva and salivary mucins: an overview. In Oral, Mucosal Diseases: Biology, Etiology and Therapy. In Proceeding of 2nd Dows Symposium, October 6–9, 1985. University of Iowa, Iowa City, Iowa (I. C. Mackenzie, C. A. Squier, and E. Dabelsteen, eds.). Copenhagen: Laegeforeningens Forlag, 1987, pp. 24–27. C. Dawes, Factors influencing salivary flow rate and composition. In Saliva and Oral Health, 2nd ed. (W. M. Edgar and D. M. O’Mullane, eds.). London: British Dental Journal, 1996, pp. 27–41. D. I. Hay, The interaction of human parotid salivary proteins with hydroxyapatite. Arch. Oral Biol. 18:1517–1529, 1973. J. L. Jensen, M. S. Lamkin, and F. G. Oppenheim, Adsorption of human salivary proteins to hydroxyapatite: a comparison between whole saliva and glandular salivary secretions. J. Dent. Res. 71:1569–1576, 1992.


844

Arnebrant

10.

M. J. Levine, Salivary macromolecules: a structure/function synopsis. In Saliva as a Diagnostic Fluid (D. Malamud and L. A. Tabak, eds.). New York: Annals of the New York Academy of Sciences, Vol. 694, 1993, pp. 11–16. M. J. Levine, M. S. Reddy, L. A. Tabak, R. E. Loomis, E. J. Bergey, P. C. Jones, R. E. Cohen, M. W. Stinson, and I. Al-Hashimi, Structural aspects of salivary glycoproteins. J. Dent. Res. 66:436–441, 1987. D. I. Hay, The adsorption of salivary proteins by hydroxyapatite and enamel. Arch. Oral Biol. 12:937–946, 1967. L. Vroman and A. L. Adams, Identification of rapid changes in plasma–solid interfaces. J. Biomed. Mater. Res. 3:43, 1969. L. Vroman, A. L. Adams, M. Klings, G. C. Fischer, P. C. Munoz, and R. P. Solensky, Reactions of formed elements of blood with plasma proteins at interfaces. Ann. NY Acad. Sci. 288:65–75, 1977. L. Vroman and A. L. Adams, Why plasma proteins interact at interfaces. In Proteins at Interfaces—Physicochemical and Biochemical Studies (J. L. Brash and T. A. Horbett, eds.). Washington, DC: American Chemical Society, 1987, pp. 154–164. J. L. Brash, Protein adsorption at the solid–solution interface in relation to blood– material interaction. In Proteins at Interfaces—Physicochemical and Biochemical Studies (J. L. Brash and T. A. Horbett, eds.). American Chemical Society, ACS Symposium Series, Vol. 343, 1987, pp. 490–506. U. Lendenmann, J. Grogan, and F. G. Oppenheim, Saliva and dental pellicle—a review. Adv. Dent. Res. 14:22–28, 2000. Y. Yao, J. Grogan, M. Zehnder, U. Lendenmann, B. Nam, Z. Wu, C. E. Costello, and F. G. Oppenheim, Compositional analysis of acquired enamel pellicle by mass spectrometry. Arch. Oral Biol. 46:293–303, 2001. Y. Yao, M. S. Lamkin, and F. G. Oppenheim, Pellicle precursor proteins: acidic prolinerich proteins, statherin, and histatins, and their crosslinking reaction by oral transglutaminase. J. Dent. Res. 78:1696–1703, 1999. L. C. P. M. Schenkels, E. C. I. Veerman, and A. V. N. Nieuw Amerongen, Biochemical composition of human saliva in relation to other mucosal fluids. Crit. Rev. Oral Biol. Med. 6:161–175, 1995. T. Imoto, L. N. Johnson, A. C. T. North, D. C. Phillips, and J. A. Rupley, Vertebrate lysozymes. In The Enzyme (P. D. Boyer, ed.). Academic Press, 1972, pp. 665–868. P. M. Claesson, E. Blomberg, J. C. Fro¨berg, T. Nylander, and T. Arnebrant, Protein interactions at solid surfaces. Adv. Colloid Interface Sci. 57:161–227, 1995. M. C. Wahlgren, T. Arnebrant, and I. Lundstro¨m, Adsorption of lysozyme to hydrophilic silicon oxide surfaces. J. Colloid Interface Sci. 175:506–514, 1995. K. H. Eggen and G. Rolla, Further information on the composition of the acquired enamel pellicle. In Cariology Today (B. Guggenheim, ed.). Karger, 1984, pp. 109–118. M. Johnsson, M. J. Levine, and G. H. Nancollas, Hydroxyapatite binding domains in salivary proteins. Crit. Rev. Oral Biol. Med. 4:371–378, 1993. D. I. Hay, J. M. Ahern, S. K. Schuckebier, and D. H. Schlesinger, Human salivary acidic proline-rich protein polymorphisms and biosynthesis studied by high-performance liquid chomatography. J. Dent. Res. 73:1717–1726, 1994. T. Li, P. Bratt, A. P. Jonsson, M. Ryberg, I. Johansson, W. J. Griffiths, T. Bergman, and N. Stro¨mberg, Possible release of an ArgGlyArgProGln Pentapeptide with innate immunity properties from acidic proline-rich proteins by proteolytic activity in commensal Streptococcus and Actinomyces species. Infect. Immun. 68:5425–5429, 2000. E. C. Moreno, K. Varughese, and D. I. Hay, Effect of human salivary proteins on the precipitation kinetics of calcium phosphate. Calcified Tissue Int. 28:7–16, 1979. D. I. Hay and E. C. Moreno, Statherin and the proline-rich proteins. In Human Saliva: Clinical Chemistry and Microbiology (J. O. Tenovou, ed.). CRC Press, 1989, pp. 131– 150.

11.

12. 13. 14.

15.

16.

17. 18.

19.

20.

21. 22. 23. 24. 25. 26.

27.

28. 29.


Protein Adsorption in the Oral Environment 30.

31. 32.

33.

34.

35.

36.

37. 38.

39.

40. 41.

42.

43.

44.

45.

46. 47.

845

W. N. Eigel, J. E. Butler, C. A. Ernstrom, H. M. Farrell, V. R. Harwalkar, R. Jenness, and R. Whitney, Nomenclature of proteins in cow’s milk. J. Dairy Sci. 67:1599–1631, 1984. C. Holt and L. Sawyer, Primary and secondary structure of the caseins in relation to their biological role. Protein Eng. 2:251, 1988. D. H. Schlesinger and D. I. Hay, Complete covalent structure of statherin, a tyrosinerich acidic peptide which inhibits calcium phosphate precipitation from human parotid saliva. J. Biol. Chem. 252:1689–1695, 1977. W. H. Douglas, E. S. Reeh, N. Ramasubbu, P. A. Raj, K. K. Bhandary, and M. J. Levine, Statherin: a major boundary lubricant of human saliva. Biochem. Biophys. Res. Commun. 180:91–97, 1991. F. G. Oppenheim, F. M. McMillian, S. M. Levitz, R. D. Diamond, G. D. Offner, and R. F. Troxler, Histatins, a novel family of histidine-rich proteins in human parotid secretion. Isolation, characterisation, primary structure and fungistatic effects on Candida albicans. J. Biol. Chem. 263:7472–7477, 1988. P. Schu¨pbach, F. G. Oppenheim, U. Lendenmann, M. S. Lamkin, Y. Yao, and B. Guggenheim, Electron-microscopic demonstration of proline-rich proteins, statherin, and histatins in acquired enamel pellicles in vitro. Eur. J. Oral Sci. 109:60–68, 2001. A. Prakobphol, M. J. Levine, L. A. Tabak, and M. S. Reddy, Purification of a lowmolecular-weight, mucin-type glycoprotein from human submandibular-sublingual saliva. Carbohydr. Res. 108:111–122, 1982. L. A. Tabak, Structure and function of human salivary mucins. Crit. Rev. Oral Biol. Med. 1:229–234, 1990. R. E. Loomis, A. Prakobphol, M. J. Levine, M. S. Reddy, and P. C. Jones, Biochemical and biophysical comparison of two mucins from human submandibular-sublingual saliva. Arch. Biochem. Biophys. 258:452–464, 1987. L. A. Tabak, M. J. Levine, J. K. Jane, R. A. Brian, R. A. Cohen, L. D. Monte, S. Zawacki, G. H. Nancollas, A. Slomiany, and B. L. Slomiany, Adsorption of human salivary mucins to hydroxyapatite. Arch. Oral Biol. 30:423–427, 1985. G. Embery, T. G. Heaney, and J. B. Stanbury, Studies on the organic polyanionic constituents of human acquired dental pellicle. Arch. Oral Biol. 31:623–625, 1986. T. Sonju and G. Rolla, Chemical analysis of the acquired pellicle formed in two hours on cleaned human teeth in vivo. Rate of formation and amino acid analysis. Caries Res. 7:30–38, 1973. P. A. Nielsen, E. P. Bennet, H. H. Wandell, M. H. Therkildsen, J. Hannibal, and H. Clausen, Identification of a major human high molecular weight salivary mucin (MG1) as tracheobronchial mucin MUC5B. Glycobiology 7:413–419, 1997. R. F. Troxler, I. Iontcheva, F. G. Oppenheim, D. P. Nunes, and G. D. Offner, Molecular characterization of a major molecular weight mucin from human sublingual gland. Glycobiology 7:965–973, 1997. D. J. Thornton, N. Kahn, R. Mehrotra, E. Veerman, N. H. Packer, and J. K. Sheehan, Salivary mucin MG1 is comprised almost entirely of different glycosylated forms of the MUC5B gene product. Glycobiology 9:293–302, 1999. R. Mehrotra, D. J. Thornton, and J. K. Sheehan, Isolation and physical characterization of the MUC7 (MG2) mucin from saliva: evidence for self-association. Biochem. J. 334:415–422, 1998. L. A. Tabak, In defence of the oral cavity: structure, biosynthesis, and function of salivary mucins. Annu. Rev. Physiol. 57:547–564, 1995. J. Brange, B. Skelbaek-Pedersen, L. Langkjaer, U. Damgaard, H. Ege, S. Havelund, L. G. Heding, K. H. Jørgensen, J. Lykkeberg, J. Markussen, et al. (eds.), Galenics of Insulin. The Physico-Chemical and Pharmaceutical Aspects of Insulin and Insulin Preparations. Berlin: Springer-Verlag, 1987.


846

Arnebrant

48.

D. G. Schmidt and T. A. J. Payens, The evaluation of positive and negative contributions to the second viral coefficient of some milk proteins. J. Colloid Interface Sci. 39: 655–662, 1972. T. A. J. Payens and B. W. Markwijk, Biochim. Biophys. Acta 71:517, 1963. P. Walstra and R. Jenness, Dairy Chemistry and Physics. New York: John Wiley and Sons, 1984. M. Rykke, G. Smistad, G. Ro¨lla, and J. Karlsen, Micelle-like structures in human saliva. Colloids Surf. B: Biointerfaces 4:33–44, 1995. M. Rykke, D. G. Devold, G. Smistad, and G. E. Vegarud, Comparison of bovine casein micelles and human salivary micelle-like structures. Int. Dairy J. 9:365–366, 1999. I. Iontcheva, F. G. Oppenheim, and R. F. Troxler, Human salivary mucin MG1 selectively forms heterotypic complexes with amylase, proline-rich proteins, statherin, and histatins. J. Dent. Res. 76:734–743, 1997. I. Iontcheva, F. G. Oppenheim, G. D. Offner, and R. F. Troxler, Molecular mapping of statherin- and histatin-binding domains in human salivary mucin MG1 (MUC5B) by the yeast two-hybrid system. J. Dent. Res. 79:732–739, 2000. A. R. Biesbrock, M. S. Reddy, and M. J. Levine, Interaction of a salivary mucin– secretory immunoglobulin A complex with mucosal antigens. Infect. Immun. 59:3492– 3497, 1991. C. Wickstrom, C. Christersson, J. R. Davies, and I. Carlstedt, Macromolecular organization of saliva: identification of ‘insoluble’ MUC5B assemblies and non-mucin proteins in the gel phase. Biochem. J. 351:421–428, 2000. M. N. Hatton, R. E. Looomis, M. I. Levine, and L. A. Tabak, Masticatory lubrication. The role of carbohydrate in the lubricating property of a salivary glycoprotein–albumin complex. Biochem. J. 230:817–820, 1985. Y. Yao, M. S. Lamkin, and F. G. Oppenheim, Pellicle precursor protein crosslinking: characterisation of an adduct between acidic proline-rich protein (prp-1) and statherin generated by transglutaminase. J. Dent. Res. 79:930–938, 2000. P.-O. Glantz, S. E. Friberg, S. M. Wirth, and R. E. Baier, Thin-section transmission electron microscopy of human saliva. Acta Odont. Scand. 47:111–116, 1989. P.-O. Glantz, J. R. Natiella, C. D. Vaughan, A. E. Meyer, and R. E. Baier, Structural studies of human saliva. Acta Odont. Scand. 47:17–24, 1989. P.-O. Glantz, S. M. Wirth, R. E. Baier, and J. E. Wirth, Electron microscopic studies of human mixed saliva. Acta Odont. Scand. 47:7–15, 1989. P. O. Glantz, M. A. Meenaghan, K. Hyun, and S. Wirth, On the presence and localization of epidermal and nerve growth factors in human whole saliva. Acta Odont. Scand. 47:287–292, 1989. P.-O. Glantz, S. E. Friberg, C. Christersson, and R. E. Baier, Surface and colloid chemical aspects of saliva particle interactions. J. Oral Rehab. 22:585–588, 1995. G. Rolla and M. Rykke, Evidence for presence of micelle-like protein globules in human saliva. Colloids Surf. B: Biointerfaces 3:117–182, 1994. M. Rykke, A. Young, G. Rolla, D. G. Devold, and G. Smistad, Transmission electron microscopy of human saliva. Colloids Surf. B: Biointerfaces 9:257–267, 1997. M. Rykke, A. Young, G. Smistad, G. Rolla, and J. Karlsen, Zeta potentials of human salivary micelle-like structures. Colloids Surf. B: Biointerfaces 6:51–56, 1996. A. Young, G. Smistad, J. Karlsen, G. Rolla, and M. Rykke, The effect of some proteolytic enzymes on human salivary micelle-like structures. Colloids Surf. B: Biointerfaces 8:189–198, 1997. J. Sjollema, H. J. Busscher, and A. H. Weerkamp, Real time enumeration of adhering microorganisms in a parallel plate flow cell using automated image analysis. J. Microbiological Methods 9:73–78, 1989.

49. 50. 51. 52. 53.

54.

55.

56.

57.

58.

59. 60. 61. 62.

63. 64. 65. 66. 67.

68.


Protein Adsorption in the Oral Environment 69. 70.

71.

72. 73. 74.

75. 76. 77.

78. 79. 80. 81.

82.

83.

84. 85. 86. 87. 88.

89.

90.

847

W. G. Pitt, M. O. McBride, A. J. Barton, and D. Sagers, Air–water interface displaces adsorbed bacteria. Biomaterials 14:605–608, 1993. C. Gomez-Suarez, H. C. van der Mei, and H. J. Busscher, Air bubble–induced detachment of polystyrene particles with different sizes from collector surfaces in a parallel plate flow chamber. Colloids Surf. A 186:211–219, 2001. C. Gomez-Suarez, H. J. Busscher, and H. C. van der Mei, Analysis of bacterial detachment from substratum surfaces by the passage of air/liquid interfaces. Appl. Environ. Microbiol. 67:2531–2537, 2001. J. Sefton, T. Arnebrant, and P.-O. Glantz, Initial studies on the behaviour of salivary proteins at liquid/air interfaces. Acta Odont. Scand. 50:221–226, 1992. E. Adamczyk, T. Arnebrant, and P.-O. Glantz, Time-dependent interfacial tension of whole saliva and saliva–bacteria mixes. Acta Odont. Scand. 55:384–389, 1998. M. L. Wallwork, J. Kirkham, J. Zhang, D. A. Smith, S. J. Brookes, R. C. Shore, S. R. Wood, O. Ryu, and C. Robinson, Binding of matrix proteins to developing enamel crystals: an atomic force microscopy study. Langmuir 17:2508–2513, 2001. J. Arends and W. L. Jongebloed, The enamel substrate—characteristics of the enamel surface. Swed. Dent. J. 1:215–224, 1977. M. Finke, K. L. Jandt, and D. M. Parker, The early stages of native enamel dissolution studied with atomic force microscopy. J. Colloid Interface Sci. 232:156–164, 2000. J. Kirkham, J. Zhang, S. J. Brookes, R. C. Shore, S. R. Wood, D. A. Smith, M. L. Wallwork, O. Ryu, and C. Robinson, Evidence for charge domains on developing enamel crystal surfaces. J. Dent. Res. 79:1943–1947, 2000. S. Chandler and D. W. Fuerstenau, In Adsorption On and Surface Chemistry of Hydroxy Apatite (D. N. Misra, ed.). New York: Plenum Press, 1984, p. 29. J. Arends, Zeta potentials of enamel and apatites. J. Dent. 7:246–253, 1979. M. E. Neiders, L. Weiss, and T. L. Cudney, An electrokinetic characterization of human tooth surfaces. Arch. Oral Biol. 15:135–151, 1970. J. Olsson, A. Carleń, and K. Holmberg, Inhibition of Streptococcus mutans adherence to hydroxy apatite with combination of alkyl phosphates and nonionic surfactants. Caries Res. 25:51–57, 1991. I. C. Hahn Berg, U. M. Elofsson, A. M. M. Joiner, and T. Arnebrant, Salivary protein adsorption onto hydroxyapatite and SDS-mediated elution studied by in situ ellipsometry. Biofouling 17:173–187, 2001. A. H. Weerkamp, H. M. Uyen, and H. J. Busscher, Effect of zeta potential and surface energy on bacterial adhesion to uncoated and saliva-coated human enamel and dentin. J. Dent. Res. 67:1483–1487, 1988. H. J. Busscher, D. H. Retief, and J. Arends, Relationship between surface free energies of dental resins and bond strengths to etched enamel. Dent. Mater. 3:60–63, 1987. M. Aning, D. O. Welch, and B. S. H. Royce, The surface energy of fluorapatite. Phys. Lett. 37A:253–254, 1971. P.-O. Glantz, On wettability and adhesiveness. Odontol. Rev. 20:1–132, 1969. R. E. Baier and R. C. Dutton, Initial events in interactions of blood with a foreign surface. J. Biomed. Mater. Res. 3:191, 1969. H. J. Busscher, J. Engelberg, and H. C. van der Mei, Initial microbial adhesion is a determinant for the strength of biofilm adhesion. FEMS Microbiol. Lett. 128:229–234, 1995. M. N. Hatton, M. J. Levine, J. E. Margarone, and A. Aguirre, Lubrication and viscosity features of human saliva and commercially available saliva substitutes. J. Oral Maxillofac. Surg. 45:496–499, 1987. C. Christersson, L. Lindh, and T. Arnebrant, Film-forming properties and viscosities of saliva substitutes and human whole saliva. Eur. J. Oral Sci. 108:418–425, 2000.


848

Arnebrant

91.

A. Aquirre, B. Mendoza, M. S. Reddy, F. A. Scannapieco, M. J. Levine, and M. N. Hatton, Lubrication of selected salivary molecules and artificial salivas. Dysphagia 4: 95–100, 1989. J. Mellema, H. J. Holterman, H. A. Waterman, and C. Blom, Rheological aspects of mucin-containing solutions and saliva substitutes. Biorheology 29:231–249, 1992. W. A. van der Reijden, E. Veerman, and A. V. N. Nieuw Amerongen, Rheological properties of commercially available polysaccharides with potential use in saliva substitutes. Biorheology 31:631–642, 1994. A. Joiner, N. M. Jones, and S. J. Raven, Investigation of factors influencing stain formation utilising an in situ model. Adv. Dent. Res. 9:471–476, 1995. R. E. Baier and P.-O. Glantz, Characterization of oral in vivo films formed on different types of solid surfaces. Acta Odont. Scand. 36:289–301, 1978. N. Vassilakos, P.-O. Glantz, and T. Arnebrant, Reflectometry—a simple optical method for investigations of salivary film formation in the oral cavity. Scand. J. Dent. Res. 101:339–343, 1993. M. Hannig, Ultrastructural investigation of pellicle morphogenesis at two different intraoral sites during a 24 h period. Clin. Oral Invest. 3:88–95, 1999. K. K. Skjørland, M. Rykke, and T. Sønju, Rate of pellicle formation in vivo. Acta Odont. Scand. 53:358–362, 1995. N. Vassilakos, T. Arnebrant, and P.-O. Glantz, Adsorption of whole saliva onto hydrophilic and hydrophobic solid surfaces. The influence of concentration, ionic strength and pH. Scand. J. Dent. Res. 100:346–353, 1992. N. Vassilakos, T. Arnebrant, and P.-O. Glantz, An in vitro study of salivary film formation at solid/liquid interfaces. Scand. J. Dent. Res. 101:133–137, 1993. L. Lindh, T. Arnebrant, P.-E. Isberg, and P.-O. Glantz, Concentration dependence of adsorption from human whole resting saliva at solid/liquid interfaces—an ellipsometric study. Biofouling 14:189–196, 1999. T. Nylander, T. Arnebrant, and P.-O. Glantz, Interactions between salivary films adsorbed on mica surfaces. Colloids Surf. A: Physicochem. Eng. Aspects 129–130:339– 344, 1997. T. Ericson, K. M. Pruitt, H. Arwin, and I. Lundstro¨m, Ellipsometric studies of film formation on tooth enamel and hydrophilic silicon surface. Acta Odont. Scand. 40: 197–201, 1982. H. J. Trurnit, Studies of enzyme systems at solid–liquid interface. II. The kinetics of adsorption and reaction. Arch. Biochem. Biophys. 51:176–199, 1954. J. M. M. Kop, J. W. Corsel, M. P. Janssen, P. A. Cuypers, and W. T. Hermens, Ellipsometric measurements of the association of prothrombine with phospholipid monolayers. J. Physique C10:491–493, 1983. L. Lindh, P.-O. Glantz, P.-E. Isberg, and T. Arnebrant, An in vitro study of initial adsorption from human parotid and submandibular/sublingual resting saliva at solid/ liquid interfaces. Biofouling 17:227–239, 2001. H. J. Busscher, M. van der Kuijl, J. Haker, R. Kalicharan, H. C. van der Mei, E. C. I. Veerman, and A. V. Nieuw Amerongen, Selective adsorption of salivary mucins by immunogold labeling. J. Dent. Res. 71:601, 1992. L. Lindh, P.-O. Glantz, I. Carlstedt, C. Wickstrom, and T. Arnebrant, Adsorption of MUC5B and the role of mucins in early salivary film formation. Colloid Surf. B: Biointerfaces 2001 (in press). L. Vroman and A. L. Adams, Adsorption of proteins out of plasma and solution in narrow spaces. J. Colloid Interface Sci. 111:391–402, 1986. T. Nylander and N. M. Wahlgren, Competitive and sequential adsorption of ␤-casein and ␤-lactoglobulin on hydrophobic surfaces and the interfacial structure of ␤-casein. J. Colloid Interface Sci. 162:151–162, 1994.

92. 93.

94. 95. 96.

97. 98. 99.

100. 101.

102.

103.

104. 105.

106.

107.

108.

109. 110.



112.

113.

114. 115. 116.

117.

118.

119.

120. 121. 122.

123.

124.

125.

126. 127.

128. 129.

849

E. Dickinson, B. S. Murray, and G. Stainsby, Protein adsorption at the air–water and oil–water interfaces. In Advances in Food Emulsions and Foams (E. Dickinson and G. Stainsby, eds.). London: Elsevier Applied Science, 1988, pp. 123–162. D. Horsley, J. Herron, V. Hlady, and J. D. Andrade, Human hen lysozyme adsorption: a comparative study using total internal reflection fluorescence spectroscopy and molecular graphics. In Proteins at Interfaces—Physicochemical and Biochemical Studies (J. L. Brash and T. A. Horbett, eds.). American Chemical Society, 1987, pp. 290–305. ACS Symposium Series, Vol. 343. J. L. Bohnert, T. A. Horbett, B. D. Ratner, and F. H. Royce, Adsorption of proteins from artificial tear solution to contact lens materials. Invest. Ophtamol. Visual Sci. 29: 362–373, 1988. U. Jo¨nsson, B. Ivarsson, I. Lundstro¨m, and L. Berghem, Adsorption behavior of fibronectin on well-characterized silica surfaces. J. Colloid Interface Sci. 90:148–163, 1982. M. Malmsten, N. Burns, and A. Veide, Electrostatic and hydrophobic effects of oligopeptide insertions on protein adsorption. J. Colloid Interface Sci. 204:104–111, 1998. N. Vassilakos, Some biophysical aspects of salivary film formation. Studies of salivary adsorption at solid/liquid and air/liquid interfaces. Ph.D. thesis. Lund University, Lund, Sweden, 1992. N. Vassilakos, T. Arnebrant, and P.-O. Glantz, Interaction of anionic and cationic surfactants with salivary pellicles formed at solid surfaces in vitro. Biofouling 5:277–286, 1992. N. Vassilakos, T. Arnebrant, and J. Rundegren, In vitro interactions of delmopinol hydrochloride with salivary films adsorbed at solid/liquid interfaces. Caries Res. 27: 176–182, 1993. N. Vassilakos, T. Arnebrant, J. Rundegren, and P.-O. Glantz, In vitro interactions of anionic and cationic surfactants with salivary fractions on well defined solid surfaces. Acta Odont. Scand. 50:179–188, 1992. N. Vassilakos, J. Rundegren, T. Arnebrant, and P.-O. Glantz, Adsorption from salivary fractions at solid/liquid and air/liquid interfaces. Arch. Oral Biol. 37:549–557, 1992. T. Arnebrant and T. Simonsson, The effect of ionic surfactants on salivary proteins adsorbed on silica surfaces. Acta Odont. Scand. 49:281–288, 1991. L. Lindh, P.-O. Glantz, N. Stro¨mberg, and T. Arnebrant, On the adsorption of human acidic proline rich proteins (PRP-1 and PRP-3) and statherin at solid/liquid interfaces. Biofouling 18:87–94, 2002. I. C. Hahn Berg, S. M. M. Kalfas, and T. Arnebrant, Proteolytic degradation of oral biofilms in vitro and in vivo: potential of proteases originating from Euphausia superba for plaque control. Eur. J. Oral Sci. 109:316–324, 2001. T. Ericson and H. Arwin, Molecular basis of saliva-mediated aggregation. In Molecular Basis of Oral Microbial Adhesion (S. E. Mergenhagen and B. Rosan, eds.). Washington, DC: American Society for Microbiology, 1985, pp. 144–150. R. E. Baier, A. E. Meyer, D. M. Dombroski, U. Nassar, J. M. Merrick, D. I. Hay, and M. P. Olivieri, Surface expression of preferential bacterial binding sites by adsorbed salivary molecules. Surfaces in Biomaterials 1:11–22, 1994. C. Christersson and R. Dunford, Salivary film formation on defined solid surfaces in the absence and presence of microorganisms. Biofouling 3:237–250, 1991. C. E. Christersson, On salivary film formation and bacterial retention to solids. A methodological and experimental in vitro study. Thesis, Lund University, Lund, Sweden, 1991. P.-O. Glantz and R. E. Baier, Recent studies of nonspecific aspects of intraoral adhesion. J. Adhesion 20:227–244, 1986. P.-O. Glantz, R. E. Baier, and C. E. Christersson, Biochemical and physiological considerations for modelling biofilms in the oral cavity. Dent. Mater. 12:208–214, 1996.


850

Arnebrant

130.

D. I. Hay, Some observations on human saliva proteins and their role in the formation of the acquired enamel pellicle. J. Dent. Res. 48:806–810, 1969. D. I. Hay, Arch. Oral Biol. 20:553–558, 1975. D. I. Hay and E. C. Moreno, Differential adsorption and chemical affinities of proteins for apatitic surfaces. J. Dent. Res. (Special Issue B) 58:930–942, 1979. E. C. Moreno, M. Kresak, and D. I. Hay, Adsorption of two human parotid salivary macromolecules on hydroxy-, fluorhydroxy-, and fluorapatites. Arch. Oral Biol. 23: 525–533, 1978. E. C. Moreno, M. Kresak, and D. I. Hay, Adsorption thermodynamics of acidic prolinerich human salivary proteins onto calcium apatites. J. Biol. Chem. 257:2981–2989, 1982. B. Lassen, K. Holmberg, C. Brink, A. Carleń, and J. Olsson, Binding of salivary proteins and oral bacteria to hydrophobic and hydrophilic surfaces in vivo and in vitro. Colloid Polym. Sci. 272:1143–1150, 1994. H. Elwing, S. Welin, A. Askendal, U. Nilsson, and I. Lundstro¨m, A wettability gradient method for studies of macromolecular interactions at the liquid/solid interface. J. Colloid Interface Sci. 119:203–209, 1987. H. J. Holterman, E. J. Gravenmade, H. A. Waterman, J. Mellema, and C. Blom, An ellipsometric study of protein adsorption at the saliva–air interface. J. Colloid Interface Sci. 128:523–532, 1989. W. Norde, Adsorption of proteins from solutions at the solid liquid interface. Adv. Colloid Interface Sci. 25:267–340, 1986. W. Norde, F. MacRitchie, G. Nowicka, and J. Lyklema, Protein adsorption at solid– liquid interfaces: reversibility and conformational aspects. J. Colloid Interface Sci. 112: 447–456, 1986. T. A. Horbett and J. L. Brash, Proteins at interfaces: current issues and future prospects. In Proteins at Interfaces—Physicochemical and Biochemical Studies (J. L. Brash and T. A. Horbett, eds.). American Chemical Society, 1987, pp. 1–37. ACS Symposium Series, Vol. 343. M. Wahlgren and T. Arnebrant, Adsorption of ␤-lactoglobulin into silica, methylated silica and polysulphone. J. Colloid Interface Sci. 136:259–265, 1990. M. Wahlgren and T. Arnebrant, Protein adsorption to solid surfaces. TIBTECH 9:201– 208, 1991. R. E. Baier, A. E. Meyer, J. R. Natiella, R. R. Natiella, and J. M. Carter, Surface properties determine bioadhesive outcomes: methods and results. J. Biomed. Mater. Res. 18:337–355, 1984. M. Quirynen, M. Marechal, H. J. Busscher, A. H. Weerkamp, J. Arends, P. L. Darius, and D. vanSteenberghe, The influence of surface free-energy on planimetric plaque growth in man. J. Dent. Res. 68:796–799, 1989. J. L. Bohnert and T. A. Horbett, Changes in adsorbed fibrinogen and albumin interaction with polymers indicated by decreases in detergent elutability. J. Colloid Interface Sci. 111:363–377, 1986. A. Carleń, A.-C. Bo¨rjesson, K. Nikdel, and J. Olsson, Composition of pellicles formed in vivo on tooth surfaces in different parts of the dentition, and in vitro on hydroxyapatite. Caries Res. 32:447–455, 1998. A. Carleń, K. Nikdel, A. Wennerberg, K. Holmberg, and J. Olsson, Surface characteristics and in vitro biofilm formation on glass ionomer and composite resin. Biomaterials 22:481–487, 2001. E. Perez and J. E. Proust, Forces between mica surfaces covered with adsorbed mucin across aqueous solution. J. Colloid Interface Sci. 118:182–191, 1987. M. Malmsten, E. Blomberg, P. M. Claesson, I. Carlstedt, and I. Ljusegren, Mucin layers on hydrophobic surfaces studied with ellipsometry and surface force measurements. J. Colloid Interface Sci. 151:579–590, 1992.

131. 132. 133.

134.

135.

136.

137.

138. 139.

140.

141. 142. 143.

144.

145.

146.

147.

148. 149.


Protein Adsorption in the Oral Environment 150. 151. 152.

153. 154.

155. 156.

157.

158.

159. 160. 161.

162. 163. 164. 165.

166. 167.

168. 169. 170.

851

G. Bernardi and T. Kawasaki, Chromatography of polypeptides and proteins on hydroxy apatite columns. Biochim. Biophys. Acta 160:301–310, 1968. T. Ericson, Adsorption to hydroxyapatite of proteins and conjugated proteins from human saliva. Caries Res. 1:52–58, 1967. T. Ericson, Salivary glycoproteins: composition and adsorption to hydroxyapatite in relation to the formation of dental pellicles and calculus. Acta Odont. Scand. 26:3–21, 1968. E. C. Moreno, M. Kresak, and D. I. Hay, Adsorption of salivary proteins onto Ca apatites. Biofouling 4:3–24, 1991. T. Nylander, T. Arnebrant, R. E. Baier, and P.-O. Glantz, Interactions between layers of salivary acidic proline rich protein 1 (PRP 1) adsorbed on mica surfaces. Prog. Colloid Polym. Sci. 108:34–39, 1998. T. Nylander and N. M. Wahlgren, Forces between adsorbed layers of ␤-casein. Langmuir 13:6219–6225, 1997. T. Nylander, F. Tiberg, T.-J. Su, and R. K. Thomas, ␤-Casein adsorption at the hydrophobized silicon oxide–aqueous solution interface and the effect of added electrolyte. Biomacromolecules 2:278–287, 2001. C. Wickstro¨m, J. R. Davies, G. V. Eriksen, E. C. I. Veerman, and I. Carlstedt, MUC5B is a major gel-forming, oligomeric mucin from human salivary gland, respiratory tract and endocervix: identification of glycoforms and C-terminal cleavage. Biochem. J. 334: 685–693, 1998. M. J. Levine, L. A. Tabak, M. Reddy, and I. R. Mandel, Nature of salivary pellicles in microbial adherence: role of salivary mucins. In Molecular Basis of Oral Microbial Adhesion (S. E. Mergenhagen and B. Rosan, eds.). Washington, DC: IRL Press, 1985, pp. 125–130. M. Malmsten, I. Ljusegren, and I. Carlstedt, Ellipsometry studies of the mucoadhesion of cellulose derivates. Colloids Surf. B: Biointerfaces 2:463–470, 1994. L. Shi and K. D. Caldwell, Mucin adsorption to hydrophobic surfaces. J. Colloid Interface Sci. 224:372–381, 2000. J. E. Proust, A. Baszkin, E. Perez, and M. M. Boissonnade, Bovine submaxillary mucin (BSM) adsorption at solid/liquid interfaces and surface forces. Colloids Surf. 10:43– 52, 1984. A. V. N. Amerongen, C. H. Oderkerk, and E. C. I. Veerman, Interaction of human salivary mucins with hydroxyapatite. J. Biol. Buccale 17:85–92, 1989. A. H. Meckel, The nature and importance of organic deposits on dental enamel. Caries Res. 2:104–114, 1968. W. G. Armstrong, The composition of organic films formed on human teeth. Caries Res. 1:89–103, 1967. B. L. Slomiany, V. L. N. Murty, E. Zdebska, A. Slomiany, K. Gwozdzinski, and I. D. Mandel, Tooth surface lipids and their role in protection of dental enamel against lactic acid diffusion in man. Arch. Oral Biol. 31:187–191, 1986. B. Larsson, G. Olivecrona, and T. Ericson, Lipids in human saliva. Arch. Oral Biol. 41:105–110, 1996. V. L. N. Murty, B. L. Slomiany, W. Laszewicz, K. Petropoulou, and I. D. Mandel, Lipids of developing dental plaque in caries-resistant and caries-susceptable adult people. Arch. Oral Biol. 30:171–175, 1985. T. Sonju, Investigation of some glycoproteins and their possible role in pellicle formation. Norske Tannlaegeforenings Tidene 85:393–403, 1975. G. Bernardi and W. H. Cook, Biochim. Biophys. Acta 44:96–105, 1960. I. Al-Hashimi and M. J. Levine, Characterization of in vivo salivary-derived enamel pellicle. Arch. Oral Biol. 34:289–295, 1989.


852

Arnebrant

171.

M. S. Lamkin, D. Migliari, Y. Yao, R. F. Troxler, and F. G. Oppenheim, New in vitro model for the acquired enamel pellicle: pellicles formed from whole saliva show intersubject consistency in protein composition and proteolytic fragmentation patterns. J. Dent. Res. 80:385–388, 2001. H. J. Busscher, H. M. W. Uyen, I. Stokroos, and W. L. Jongebloed, A transmission electron microscopy study of the adsorption patterns of early developing artificial pellicles on human enamel. Arch. Oral Biol. 34:803–810, 1989. M. Hannig, Transmission electron microscopic study of in vivo pellicle formation on dental materials. Eur. J. Oral Sci. 105:422–433, 1997. T. A. Horbett, Adsorption to biomaterials from protein mixtures. In Proteins at Interfaces—Physicochemical and Biochemical Studies (J. L. Brash and T. A. Horbett, eds.). American Chemical Society, 1987, pp. 239–260. ACS Symposium Series, Vol. 343. T. A. Horbett and J. L. Brash, Proteins at interfaces II. Fundamentals and applications. ACS Symposium, 1995. G. Embery, S. D. Hogg, T. G. Heaney, J. B. Stanbury, and D. R. J. Green, Some considerations on dental pellicle formation and early bacterial colonisation: the role of high and low molecular weight proteins of the major and minor salivary glands. In Bacterial Adhesion and Preventive Dentistry (J. M. ten Cate, ed.). Proceedings of a workshop: Bacterial adhesion to biosurfaces with special emphasis on dental problems, November 28–December 1, 1983 Paterswolde, The Netherlands. Oxford: IRL Press, 1984, pp. 73–84. W. A. Nolte, Oral Microbiology, 2nd ed. St. Louis: Mosby, 1973. H. J. Busscher and L. V. Evans (eds.), Oral Biofilms and Plaque Control. Amsterdam: Harwood Academic, 1998. J. W. Costerton and Z. Lewandowski, The biofilm lifestyle. In Biofilms on Oral Surfaces: Implications for Health and Disease. 14th international conference on oral biology, Monterey, CA (M. J. Novak, ed.). International American Associations for Dental Research, 1997, pp. 192–196. M. A. Listgarten, The structure of dental plaque. Periodontology 5:52–65, 2000. B. Rosan and R. J. Lamont, Dental plaque formation. Microbes and Infection 2:1599– 1607, 2000. T. Lie, Pellicle formation on hydroxyapatite splints attached to the human dentition: morphologic confirmation of the concept of adsorption. Arch. Oral Biol. 20:739–742, 1975. W. Norde and Lyklema, Protein adsorption and bacterial adhesion to solid surfaces: a colloid chemical approach. Colloids Surf. 38:1–13, 1989. H. J. Busscher and A. H. Weerkamp, Specific and non-specific interactions in bacterial adhesion to solid surfaces. FEMS Microbiol. Rev. 46:165–173, 1987. R. J. Gibbons, D. I. Hay, W. C. I. Childs, and G. Davis, Role of cryptic receptors (cryptitopes) in bacterial adhesion to oral surfaces. Arch. Oral Biol. 35(Suppl ):107s– 114s, 1990. P.-O. Glantz, T. Arnebrant, T. Nylander, and R. E. Baier, Bioadhesion—a phenomenon with multiple dimensions. Acta Odont. Scand. 57:238–241, 1999. P. Glantz, R. Baier, R. Attstro¨m, and A. Meyer, Comparative clinical wettability of teeth and intraoral mucosa. J. Adhesion Sci. Technol. 5:401–408, 1991. M. D. Jendresen, Studies on intraoral adhesion. An in-vivo study of clinical adhesiveness of teeth and selected dental materials. Ph.D. thesis, Lund University, Lund, Sweden, 1980. H. C. van der Mei, A. J. Leonard, A. H. Weerkamp, P. G. Rouxhet, and H. J. Busscher, Properties of oral Streptococci relevant for adherence: zeta potential surface free energy and elemental composition. Colloids Surf. 32:297–305, 1988.

172.

173. 174.

175. 176.

177. 178. 179.

180. 181. 182.

183. 184. 185.

186. 187. 188.

189.



191. 192.

193. 194.

195.

196.

197.

198. 199.

200.

201.

202.

203.

204.

205.

206. 207. 208. 209.

853

H. C. van der Mei, A. H. Weerkamp, and H. J. Busscher, A comparison of various methods to determine hydrophobic properties of streptococcal cell surfaces. J. Microbiol. Methods 6:277–287, 1987. G. Westergren and J. Olsson, Hydrophobicity and adherence of oral Streptococci after repeated subculture in vitro. Infect. Immun. 40:432–435, 1983. W. A. Zisman, Relation of the equilibrium contact angle to liquid and solid constitution. In Contact Angles, Wettability and Adhesion (R. Gould, ed.). American Chemical Society, 1964, pp. 1–51. ACS Adv. Chem. Ser., Vol. 43. R. E. Baier, E. G. Schafrin, and W. A. Zisman, Science 162:1360, 1968. R. E. Baier and A. E. Meyer, Surface energetics and biological adhesion. In International Symposium on Physicochemical Aspects of Polymer Surfaces (K. L. Mittal, ed.). New York: Plenum Press, 1981, pp. 895–909. R. J. Gibbons, D. I. Hay, J. O. Cisar, and W. B. Clark, Adsorbed salivary proline-rich protein 1 and statherin: receptors for type 1 fimbriae of Actenomyces viscosus T14VJ1 on apatitic surfaces. Infect. Immun. 56:2990–2993, 1998. A. Carleń, P. Bratt, C. Stenudd, J. Olsson, and N. Stro¨mberg, Agglutinin and acidic proline-rich protein receptor patterns may modulate bacterial adherence and colonization on tooth surfaces. J. Dent. Res. 77:81–90, 1998. R. J. Gibbons and D. I. Hay, Human salivary acidic proline-rich proteins and statherin promote the attachment of Actinomyces viscosus LY7 to apatitic surfaces. Infect. Immun. 56:439–445, 1988. R. Gibbons and D. I. Hay (eds.), Adsorbed Salivary Proline-Rich Proteins as Bacterial Receptors on Apatitic Surfaces. New York: Springer-Verlag, 1989. R. J. Gibbons and D. I. Hay, Adsorbed salivary acidic proline rich proteins contribute to the adhesion of Streptococcus mutans JBP to apatitic surfaces. J. Dent. Res. 68: 1303–1307, 1989. R. J. Gibbons, D. I. Hay, and D. H. Schlesinger, Delineation of a segment of adsorbed salivary acidic proline-rich proteins which promotes adhesion of Streptococcus gordonii to apatitic surfaces. Infect. Immun. 59:2948–2954, 1991. C. G. van Hoogmoed, M. van der Kuijl-Booij, H. C. van der Mei, and H. J. Busscher, Inhibition of Streptococcus mutans NS adhesion to glass with and without a salivary conditioning film by biosurfactant-releasing Streptococcus mitis strains. Appl. Environ. Microbiol. 66:659–663, 2000. J. W. T. Wimpenny and R. Colasanti, A unifying hypothesis for the structure of microbial biofilms based on cellular automaton models. FEMS Microbiol. Ecology 22:1– 16, 1997. I. W. Sutherland, Biofilm matrix polymers. In Dental Plaque Revisited—Oral Biofilms in Health and Disease (H. N. Newman and M. Wilson, eds.). Eastman Dental Institute, University College London and BioLine, Cardiff School of Biosciences, 1999. D. J. Hassett, J. G. Elkins, J.-F. Ma, and T. R. McDermott, Pseudomonas aeruginosa biofilm sensitivity to biocides: use of hydrogen peroxide as model antimicrobial agent for examining resistance mechanisms. Method Enzymol. 310:599–608, 1999. D. G. Davies, M. R. Parsek, J. P. Pearson, B. H. Iglewski, J. W. Costerton, and E. P. Greenberg, The involvement of cell-to-cell signals in the development of a bacterial biofilm. Science 280:295–298, 1998. R. Kolter and R. Losick, One for all and all for one. Science 280:226–227, 1998. J. W. Costerton, D. E. Lewandowski, D. R. Caldwell, D. R. Korber, and H. M. LappinScott, Microbial biofilms. Annu. Rev. Microbiol. 49:711–745, 1995. J. W. Costerton, Biofilms 2000. In Biofilms 2000 (J. W. Costerton, ed.). Washington, DC: American Society for Microbiology, 2000. N. Vassilakos, S. Kalfas, T. Arnebrant, and Rundegren, A simple flow cell system to evaluate in vitro bacterial adhesion on solids. Colloids Surf. B: Biointerfaces 1:341– 347, 1993.


854

Arnebrant

210.

A. Kharazmi, B. Giwercman, and N. Hiby, Robbins device in biofilm research. Method Enzymol. 310:207–215, 1999. M. S. Zinn, R. D. Kirkegaard, R. J. Palmer, and D. C. White, Laminar flow chamber for continuous monitoring of biofilm formation and succession. Method Enzymol. 310: 224–232, 1999. D. G. Ahearn, R. N. Borazjani, R. B. Simmons, and M. M. Gabriel, Primary adhesion of Pseudomonas aeruginosa to inanimate surfaces including biomaterials. Method Enzymol. 310:551–557, 1999. J. Molin and S. Molin, CASE: complex adaptive system ecology. Adv. Microbial Ecology 15:27–80, 1997. J. Rundegren, T. Simonsson, L. Peterson, and E. Hansson, Effect of delmopinol on the cohesion of glucan-containing plaque formed by Streptococcus mutans in a flow cell system. J. Dent. Res. 71:1792–1796, 1992. I. D. Mandel, Caries prevention—current strategies, new directions. JADA 127:1477– 1488, 1996. A. Mombelli, Antibiotics in periodontal therapy. In Clinical Periodontology and Implant Dentistry (J. Lindhe, ed.). Copenhagen: Munksgaard, 1999, pp. 488–507. D. Steinberg and M. Friedman, Dental drug-delivery devices: local and sustained release applications. Crit. Rev. Therapeutic Drug Carrier Systems 16:425–459, 1999. M. Rykke and G. Rolla, Effect of silicone oil on protein adsorption to hydroxy apatite in vitro and pellicle formation in vivo. Scand. J. Dent. Res. 98:401–411, 1990. J. Olsson, Y. van der Heijde, and K. Holmberg, Plaque formation in vivo and bacterial attachment in vitro on permanently hydrophobic and hydrophilic surfaces. Caries Res. 26:428–433, 1992. M. Rykke, G. Rolla, and T. So¨nju, Effect of sodium lauryl sulfate on protein adsorption to hydroxyapatite in vitro and on pellicle formation in vivo. Scand. J. Dent. Res. 98: 135–143, 1990. M. Rykke, G. Rolla, and T. So¨nju, Effect of pyrophosphate on protein adsorption to hydroxyapatite in vitro and on pellicle formation in vivo. Scand. J. Dent. Res. 96:517– 522, 1988. M. Rykke and G. Rolla, Effect of two organic phosphonates on protein adsorption in vitro and on pellicle formation in vivo. Scand. J. Dent. Res. 98:486–496, 1990. J. Olsson, A. Carleń, and K. Holmberg, Inhibition of Streptococcus mutans adherence by means of surface modification. J. Dent. Res. 69:1586–1592, 1990. J. Olsson, M. Hellsten, and K. Holmberg, Surface modification of hydroxyapatite to avoid bacterial adhesion. Colloid Polym. Sci. 269, 1991. J. Olsson, A. Carleń, and K. Holmberg, Modulation of bacterial binding to salivary pellicle by treatment with hydrophilising compounds. Arch. Oral Biol. 35:1378–1405, 1990. K. Holmberg and J. Olsson, Surface modification of hydroxy apatite for dental plaque inhibition. In Desk Reference of Functional Polymers (R. Arshady, ed.). American Chemical Society, 1997. P. Schupbach, J. R. Neeser, M. Golliard, M. Rouvet, and B. Guggenheim, Incorporation of caseinoglycomacropeptide and caseinophosphopeptide into the salivary pellicle inhibits adherence of mutans Streptococci. J. Dent. Res. 75:1779–1788, 1996. T. H. Grenby, A. T. Andrews, M. Mistry, and R. J. H. Williams, Dental caries— protective agents in milk and milk products: investigations in vitro. J. Dent. 29:83–92, 2001. I. Johansson, P. Bratt, D. I. Hay, S. Schluckebier, and N. Stro¨mberg, Adhesion of Candida albicans, but not Candida krusei, to salivary statherin and mimicking host molecules. Oral Microbiol. Immunol. 15:112–118, 2000.

211.

212.

213. 214.

215. 216. 217. 218. 219.

220.

221.

222. 223. 224. 225.

226.

227.

228.

229.



231.

232. 233.

234.

235.

236. 237.

238.

239.

240.

241. 242. 243. 244.

245.

246. 247.

855

M. Edgerton, P. A. Raj, and M. J. Levine, Surface-modified poly(methyl methacrylate) enhances adsorption and retains anticandidal activities of salivary histatin 5. J. Biomed. Mater. Res. 29:1277–1286, 1995. D. J. Smith, M. A. Taubman, W. F. King, S. Eida, J. R. Powell, and J. Eastcott, Immunological characteristics of a synthetic peptide associated with a catalytic domain of mutans streptococcal glucosyltransferase. Infect. Immun. 62:5470–5476, 1994. A. M. Vacca Smith and W. H. Bowen, In situ studies of pellicle formation on hydroxyapatite discs. Arch. Oral Biol. 45:277–291, 2000. T. Simonsson, T. Arnebrant, and L. Petersson, The effect of delmopinol on salivary pellicles. The wettability of tooth surfaces in vivo and bacterial cell surfaces in vitro. Biofouling 3:251–260, 1991. T. Simonsson, T. Arnebrant, L. Petersson, and E. Bondesson Hvid, Influence of delmopinol on bacterial ␨-potentials and on the colloidal stability of bacterial suspensions. Acta Odont. Scand. 49:311–316, 1991. J. Rundegren, T. Arnebrant, and L. Lindahl, Effect of whole saliva on the rheological behaviour of extracellular water-soluble glucan produced by Streptococcus mutans. Scand. J. Dent. Res. 99:484–488, 1991. S. Goldberg and M. Rosenberg, Bacterial desorption by commercial mouthwashes vs two-phase oil:water formulations. Biofouling 3:193–198, 1991. L. B. Freitas, N. Vassilakos, and T. Arnebrant, Interactions of chlorhexidine with salivary films adsorbed at solid/liquid and air/liquid interfaces. J. Periodontal Res. 28:92– 97, 1993. ¨ gaard, Effects of oral rinsing with S. M. Waaler, G. Rolla, K. K. Skjo¨rland, and B. O triclosan and sodium lauryl sulphate on dental plaque formation. Scand. J. Dent. Res. 101:192–195, 1993. S. M. Waaler, G. Rolla, and V. Kjaerheim, Triclosan containing mouthwashes—does the nature of the solvent influence their clinical effect. Scand. J. Dent. Res. 102:46– 49, 1994. G. Rolla, J. E. Ellingsen, and D. Gaare, Polydimetylsiloxane as a tooth surface–bound carrier of triclosan: a new concept in chemical plaque inhibition. Adv. Dent. Res. 8: 272–277, 1994. C. M. McCullagh, L. M. Soby, A. M. Jamieson, and J. Blackwell, Viscoelastic behaviour of fractionated ovine submaxillary mucins. Biopolymers 32:1665–1674, 1992. E. S. Reeh, W. H. Douglas, and M. J. Levine, Lubrication of saliva substitutes at enamel–enamel contacts in an artificial mouth. J. Prosthetic Dent. 75:649–656, 1996. M. J. Levine, Development of artificial salivas. Crit. Rev. Oral Biol. Med. 4:279–286, 1993. M. C. F. Johnsson, F. A. Richardson, F. A. Scannapieco, M. J. Levine, and G. H. Nancollas, The influence of salivary proteins on the growth, aggregation and surface properties of hydroxyapatite particles. In Material Synthesis Utilising Biological Processes (P. C. Rieke, P. D. Calvert, and M. Alper, eds.). 1990, pp. 81–86. M. Johnsson, F. A. Richardson, E. J. Bergey, M. J. Levine, and G. H. Nancollas, The effects of human salivary cystatins and statherin on hydroxyapatite crystallisation. Arch. Oral Biol. 36:631–636, 1991. V. Hlady and H. Fu¨redi-Milhofer, Adsorption of human serum albumin on precipitated hydroxyapatite. J. Colloid Interface Sci. 69:460–468, 1979. H. J. Busscher and H. C. van der Mei, Physico-chemical interactions in initial microbial adhesion and relevance for biofilm formation. Adv. Dent. Res. 11:24–32, 1997.


30 Adsorbed Biopolymers: Behavior in Food Applications DAVID S. HORNE

Hannah Research Institute, Ayr, Scotland

J. M. RODRIGUEZ PATINO

I.

University of Seville, Seville, Spain

INTRODUCTION

The protein film adsorbed at the oil–water or air–water interface is the source of many of the unique properties of food emulsions and foams, particularly their stability and interactions, which translate into the shelf-life and textural properties so desired by manufacturers and appreciated by consumers. This is the intuitively attractive picture that the reviews and textbooks would have us believe but the reality is that, despite all the research effort put into studying those interactions, we still do not have a reliable predictor of shelf-life in emulsion products. As for texture, we still have no quantitative instrumental measures of creaminess, mouth-feel, or even perceived thickness. Partly this arises because the interfacial layer in food emulsions is compositionally and structurally complex, but this facile explanation conceals an underlying lack of ability to formulate a realistic interaction potential between emulsion droplets or foam bubbles. That notwithstanding, considerable progress has been made in the last 10 to 15 years in extending our knowledge of the interactions between and within adsorbed protein films, much of it on proteins of interest to the food manufacturer, and much of it reviewed in extenso by Dickinson [1–6]. The problems posed by the complexity of food systems should not, however, be underestimated. Food dispersions are complicated multicomponent systems containing many emulsifiers which may show surface activity by themselves (proteins and low-molecular-weight surfactants) or by association with other components (proteins, low-molecular-weight surfactants, and polysaccharides, to name a few examples). In addition, food dispersions contain many other organic (ethanol, sugars, etc.) and inorganic (salts) reagents which may interact with emulsifiers in different complex fashion depending on pH, temperature, processing history, etc., all of which intensify the problems of the manufacturer who is attempting to control stability, shelf-life, or product texture. For up-to-date reviews, readers are directed to recent references [7–13]. Manufacturers employ two types of emulsifier or foaming agents in foods [1]. These are low-molecular-weight surfactants (mainly mono- and diglycerides, phospholipids, etc.) and macromolecules (proteins and certain hydrocolloids). In the main, the proteins are those of milk and eggs. Because early manufacturing practices tended © 2003 by Marcel Dekker, Inc.

858

Horne and Rodriguez Patino

to mimic processes developed initially in the kitchen, recipes have not strayed far from those early listings of ingredients. Since these recipes often employed whole milk or egg, proteins and lipids frequently coexist in manufactured emulsions and foams. Sometimes they are dissociated from one another, but in other cases, where they are deliberately introduced to develop specific functions in the finished product, they are found associated [14]. Lipids stabilize the dispersed droplets or bubbles by formation of a densely packed but much less rigid monomolecular layer, which is stabilized by dynamic processes (i.e., Gibbs–Marangoni effect). Polar lipids adsorb strongly to fluid–fluid interfaces giving close molecular packing at the interface to produce low surface and interfacial tensions [15,16]. In contrast, proteins act as polymeric emulsifiers with multiple anchoring sites at the interface that, together with the unfolding process of the adsorbing protein molecule [17–19], stabilize the interfacial layer kinetically. This behavior contributes significantly to the interfacial rheological properties and immobilizes proteins in the adsorbed layer. Proteins and lipids have an important physical property in common, their amphiphilic nature. This property provides the potential for association, adsorption, and reorientation at fluid–fluid interfaces, depending on the properties of the components and the protein–lipid ratio [15,16,20,21]. However, more important in some products is the effect of the small molecule emulsifiers in destabilizing the emulsion [22,23]. In the formulation of ice cream the small molecule emulsifier (typically, mono- and diglycerides) is added to break the adsorbed layer of protein and allow the adsorption of fat to the surface of the air bubble. Thus, an important action of the small molecule emulsifiers is to promote the displacement of proteins (mainly caseins) from the interface. Competitive adsorption and/or displacement between lipids and proteins at fluid–fluid interfaces have been studied in detail in several investigations. For further information concerning the interfacial characteristics of food emulsifiers (proteins and lipids), the reader is referred to recent reviews [6,7,13,14,24–31]. This chapter will concentrate on the interfacial behavior of milk proteins. Our emphasis will be on the interface as a three-dimensional dynamic entity. Interdroplet interactions will be dominated by the dimension normal to the interface, but within the interface we will concern ourselves with the role of lateral interactions and bonding between neighboring molecules. We will consider protein structure and morphology at the interface, relaxation phenomena, and interfacial rheology at oil–water and air–water interfaces, all of this behavior relevant to the formation and stability of food emulsions and foams. Conveniently the milk proteins subdivide into two categories, the disordered flexible caseins and the compact globular whey proteins. As a first approximation, a casein monomer may be regarded as a complex linear polymer that adsorbs to an oil or air interface to give an entangled monolayer of flexible chains having some regions of more hydrophobic character in direct contact with the surface (trains) and others protruding into the aqueous phase (loops and tails) [32]. The distribution of hydrophobic residues along the casein polypeptide chain is not uniform but clustered in a fashion specific to the various family members, ␣S1-, ␣S2-, ␤ -, and ␬-casein. This gives to each of the caseins some individuality, which is further enhanced by their differing content of phosphoseryl residues. In milk, however, the caseins exist not as individual molecules but as strongly aggregated polydisperse complex particles containing all four members of the casein © 2003 by Marcel Dekker, Inc.

Adsorbed Biopolymers in Food Applications

859

family. These aggregates, known as casein micelles, are held together by a combination of hydrophobic bonding between caseins and cross-linking through calcium phosphate microcrystals bonding to the phosphoseryl clusters [33,34]. Milk owes its low fluidity to the existence of these micelles, and many related food technology uses rely on their destabilization, as in the manufacture of cheese and yogurt. Precipitating the casein micelles by reducing milk pH to 4.6 and redispersing the precipitate in NaOH at pH 7.0 produces sodium caseinate. Sodium caseinate, which contains all of the caseins in the proportions found in milk, is widely used in the food industry as an emulsifier. Much of the relevant recent research work, however, has involved two major individual caseins, ␣S1- and ␤ -casein, which together comprise 75% of total bovine milk casein. In contrast to the interfacial behavior of the caseins, research to be summarized herein has demonstrated that a closely packed globular protein network is perhaps better modelled as a dense two-dimensional network of strongly interacting particles with rheological and structural properties more akin to those of a concentrated heatset globular protein gel. Again, while there are industrial preparations of whey proteins prepared from acid or cheese wheys [whey protein isolates (WPI) and whey protein concentrates (WPC)], the majority of research has been carried out on the major constituent protein at 75% of bovine whey protein, ␤ -lactoglobulin.

II.

FILM STRUCTURE AND MORPHOLOGY

A.

␤ -Casein

In producing an emulsion, the adsorbing protein molecule is the one which happens to be in close proximity to the interface at the time it is created by high energy shearing. The protein layer effectively forms a macromolecular barrier at the oil– water interface to protect the freshly formed droplets against recoalescence [35]. While early studies concentrated on layer composition and competitive adsorption and displacement, experimental and theoretical work in recent years has emphasized the dynamic structure of the interfacial protein layer and its response to its environment. The structure of adsorbed layers of ␤ -casein at various solid and liquid interfaces has been investigated experimentally by the techniques of ellipsometry [36– 39], small angle x-ray scattering [40], dynamic light scattering [41,42], and neutron reflectivity [43–49]. Ellipsometry and dynamic light scattering provide only gross measurements of layer thickness, whereas neutron reflectivity has emerged as a technique capable of providing a more detailed density profile of the adsorbed layer on an atomic scale in a direction normal to the interface. Data analysis is still restricted to the testing and fitting of appropriate models for the adsorbed film. Although a discrete two- or multilayer model, depending on the substrate, fits the neutron reflectivity data from most studies of adsorbed ␤ -casein fairly well, a power-law model has also been employed to describe the volume fraction profile at the air–water interface [47]. Statistical modelling of the adsorption of a model ␤ -casein-like polymer at a hydrophobic planar surface has also been carried out using the self-consistent field (scf) theory of Scheutjens and Fleer [50–53]. While this approach outputs a numerical density distribution, it does not lend itself readily to manipulative data fitting. © 2003 by Marcel Dekker, Inc.

860


Only gross comparisons can therefore be made of its predictions with empirical observations. So many are the correlations with parameters drawn from all kinds of experiments that confidence in the conformations predicted from this modelling must be high. The scf calculations predict that an adsorbed ␤ -casein monolayer is produced at bulk protein concentrations below an ionic strength dependent threshold [52]. Above this threshold, there is multilayer condensation of self-associating protein onto the surface. Figure 1a shows the calculated segment density profile ␳ (z) at 10 mM ionic strength and bulk volume fraction of model protein (5 ⫻ 10⫺6) for three different pH values. Each curve shows a smooth decrease in segment density moving away from the surface. Regardless of pH, the dense inner layer is about 1 nm thick with a volume fraction, ␾, approaching 0.9. Volume fraction decreases to ⬃0.01 by z = 10 nm. As pH is reduced, additional adsorbed material appears around the middle of the profile (z ⬃ 2 nm) with a distinct shoulder appearing at pH 5.5. In Fig. 1b, we present for comparison volume fraction profiles inverted from neutron reflectivity data, measured at 20 mM ionic strength and bulk protein concentration 5 ⫻ 10⫺3 wt% [46]. These too show an increase in adsorbed protein around the midrange of the profile at pH 6.0 and a doubling of the thickness of the inner layer at pH 5.5, broadly confirming the predicted behavior. Replotting the scf data (Fig. 2a) as logarithm of the segment density as a function of distance, z, shows that the low density region (␾ ⬃ 0.01) continues out to around 15 nm before dropping abruptly by several orders of magnitude to the bulk solution volume fraction. This knee we take as defining the hydrodynamic thickness of the adsorbed layer. Inspection of this tail region shows that this hydrodynamic layer thickness is not a strong function of pH, with the calculated plots all lying very close to one another. Dynamic light scattering measurements of the hydrodynamic thickness of ␤ -casein layers adsorbed onto polystyrene latices [42] confirm this behavior with pH (Fig. 2b), showing no major collapse of the ␤ -casein layer at the reduced pH. Detailed analysis of the distribution of individual segment types can also be extracted from the scf calculations. This has shown that the most hydrophilic residues, especially the carboxyl and phosphoserine residues, reside predominantly in the outer layer regions. In particular, the N-terminal amino residue lies mainly well away from the surface. All of this is consistent with the concept of a long sterically stabilizing tail that is composed of the N-terminus region of the ␤ -casein molecule (Fig. 3). Self-consistent field calculations performed for the fully dephosphorylated model of ␤ -casein show that the loss of the phosphate residues produces an increase in surface coverage of the dephosphorylated analog, consistent with the much lower solubility of dephosphorylated ␤ -casein. The additional adsorbed material is located in the inner layers at distances 17 mN/m), and especially at the collapse point, the thickness of the ␤ -casein or caseinate monolayer is greater than that for WPI. These results are in good agreement with those obtained by atomic force microscopy [78]. Data in Fig. 8 show that a doubling in the thickness is produced at the highest surface pressure relative to the equilibrium surface pressure (␲e), the relative reflectivity increasing by about a factor of 4. At ␲e , a saturation of the monolayer takes place. For ␤ -casein at pH 7, the monolayer coverage saturation (Fig. 6) agrees well with literature data [79]. © 2003 by Marcel Dekker, Inc.


867

FIG. 6 Surface pressure–area isotherms for ␤ -casein (——), caseinate (⭈ ⭈ ⭈), and WPI monolayers (— ⭈ —) spread on aqueous solutions at pH 5 (a) and pH 7 (b) at 20⬚C.

At pH 5, the surface density of disordered proteins (␤ -casein and caseinate) was higher than that for the globular WPI (Fig. 6a). However, at pH 7 (Fig. 6b) the structural characteristics of globular (WPI) and disordered (␤ -casein and caseinate) proteins were essentially the same, especially at surface pressures lower than 17 mN/m. These differences are a consequence of the more compact packing of disordered residues in ␤ -casein and caseinate on an acidic subphase close to the isoelectric point [71]. The fact that the WPI monolayer structure did not depend on the pH (Fig. 6) is consistent with the view that the protein components of WPI (which is mainly ␤ -lactoglobulin) retains elements of the native structure, not fully unfolded, at the air–water interface.

III.

KINETICS OF ADSORPTION OF PROTEINS AT FLUID INTERFACES

The rate of protein adsorption at fluid–fluid interfaces plays an important role in the formation and stabilization of food dispersions [1,80,81]. In fact, during the formation of a dispersed system the protein must be adsorbed at the interface to prevent © 2003 by Marcel Dekker, Inc.

868


FIG. 7 Proposed structures of disordered proteins (␤ -casein) and globular proteins (␤ -lactoglobulin, BSA, etc.) at the air–water interface as a function of surface pressure or surface density. Disordered proteins adopt trains of amino acid segments at the interface (a, top), looping of the amino acid segments into the underlying aqueous solution (b, top), and finally (c, top) the collapse and multilayer formation. Globular proteins adopt loop conformation at the interface (a, bottom), which is more condensed at higher surface pressures (b, bottom), and then the collapse and multilayer formation (c, bottom). The monolayer collapse takes place at a surface pressure close to the equilibrium surface pressure.

the recoalescence of the initially formed bubbles or droplets. In addition, during the protein adsorption the surface or interfacial tension of fluid–fluid interfaces decreases, an important factor both in optimizing the energy required in the emulsification or foaming process [82] and in achieving smaller droplet and bubble size— which is an important factor for the stability of the dispersed system [1]. On the other hand, emulsification and foaming involve interfacial deformation, and the response of the adsorbed layer to such deformations is crucial for understanding the role of proteins in food systems [83]. The decrease in surface tension by proteins follows a series of different processes [84–87]. First, the protein has to move from the bulk phase to the subsurface (a layer immediately adjacent to the fluid interface) by diffusion and/or convection. This step is followed by the adsorption and unfolding of the protein at the interface. Third, the adsorbed protein segments rearrange at the fluid interface, a slow process caused by reorganization of the amino acid segments previously adsorbed on the interface. Adsorption of proteins is therefore a complex process, involving possibly © 2003 by Marcel Dekker, Inc.


869

FIG. 8 Relative intensity as a function of surface pressure for ␤ -casein (䡩), caseinate (䉭), and WPI spread films (䉮) at the air–water interface at 20⬚C and pH 5; I: 0.05 M.

several conformational changes that may be either reversible or irreversible and, in addition, may be time dependent [88,89]. When direct measurements of surface concentration, ⌫(␪ ), are not possible, the kinetics of adsorption can be monitored by measuring the changes in surface pressure, ␲, with time, for which a range of techniques have been developed [90,91]. For the long-term stages of protein adsorption, which are controlled by unfolding at the interface and configurational rearrangements of the adsorbed molecules, the Wilhelmy plate is an adequate method. However, the diffusion step is too fast to be detected with this technique [17,18,92,93] and, for measurements at short adsorption time, an automatic drop tensiometer has been employed [19,94]. The same tensiometer has been used for the analysis of WPI adsorption at the oil–water interface [28,95]. Some authors are critical of the use of surface pressure measurements following adsorption [83,96–98], as it implicitly assumes that the surface equation of state is the same for spread and adsorbed protein films. However, it was recently shown by ˜ and Rodriguez Patino [99] that the agreement between spread and Rodriguez Nino adsorbed BSA monolayers is generally good, implying that the structures of the monolayers formed in the two different ways must be identical, at least for adsorption from low bulk protein concentrations. The same agreement between the adsorbed and spread isotherms has been observed for BSA [83,100,101] and for other proteins [83,102]. The good agreement between spread and adsorbed ␲-A isotherms provides support for the relevance of surface pressure measurements to adsorption studies [83,103]. A.

Protein Adsorption at the Oil–Water Interface

For WPI adsorption at the oil–water interface we have observed that interfacial pressure, ␲, and surface dilatational modulus, E, increase; and the phase angle, ␾, © 2003 by Marcel Dekker, Inc.

870


decreases with time, ␪ (Fig. 9). Normalization into a single master curve of E versus ␲ data (Fig. 10) reflects the interfacial behavior of WPI adsorbed films for different protein concentrations, at different adsorption times, and under different processing conditions and suggests that the interfacial behavior of WPI films is mainly due to the amount of adsorbed protein. The rate of WPI adsorption at the oil–water interface increases with protein concentration in solution (Fig. 9). The kinetics at short adsorption time are controlled by the diffusion of the protein toward the interface, in agreement with the Ward and Tordai model [104]. However, at long-term adsorption, a first-order model is a sat-

FIG. 9 Time-dependent surface pressure (a), surface dilatational modulus (b), and phase angle for WPI adsorbed films at the oil–water interface (c) at pH 5 and at 20⬚C; I: 0.05 M; frequency: 100 MHz; amplitude: 15%. Protein concentration in the drop bulk phase (%wt/ wt): 10⫺1 (䡩), 10⫺2 (䉭), and 10⫺5 (▫).



871

FIG. 10 Surface dilatational modulus as a function of surface pressure for WPI adsorbed films at the oil–water interface at pH 5 and 20⬚C; I: 0.05 M; frequency: 100 MHz; amplitude: 15%. Protein concentration in the drop bulk phase (%wt/wt): 10⫺1 (䡩), 10⫺2 (䉭), and 10⫺5 (䉮).

isfactory mathematical description of the rheokinetic data, which show exponential changes in surface pressure or dilatational modulus with time [28,95]. These phenomena can be related to the protein unfolding and/or protein–protein interactions as a function of protein concentration in solution. In practice, the plot of ln ␲ or ln E versus time usually deviates from linearity. The initial slope is taken to correspond to a first-order rate constant of unfolding, while the second, later, slope is taken to correspond to a first-order rate constant of rearrangement, occurring among a more-or-less constant number of adsorbed molecules. Transient surface dynamic properties of WPI adsorbed films also depend on the WPI concentration in the bulk phase [105]. It was observed that the rate of surface tension or surface dilatational modulus change over time increased when the WPI concentration in the bulk phase was increased. At high concentrations, the surface activity and surface dilatational modulus were high. Protein adsorption at the interface was therefore facilitated at higher protein concentrations in the bulk phase. However, the measured interfacial tension or surface dilatational modulus continued to increase with time, even at long-term adsorption. That is, a steady-state value of the surface dilatational modulus was not attained even at the higher adsorption time studied here. Thus, the study of WPI adsorption is very time consuming, especially at the lower protein concentrations in the bulk phase. Over the adsorption period studied here, the film behaved, from a rheological point of view, as viscoelastic with a phase angle greater than zero. This phase angle decreased with time, with more marked time dependence at higher protein concentrations in the bulk phase. Such behavior is consistent with the existence of higherorder protein–protein interactions, which are thought to be due to a higher protein concentration at the interface [95,105], because of both the longer adsorption time and the higher protein concentration in the bulk phase. © 2003 by Marcel Dekker, Inc.

872


In these experiments we have noted that the interfacial pressure and its rate of change for WPI adsorbed film in the presence of convection are the same as in the absence of convection [95]. As reported by Damodaran [81], the interfacial adsorption of proteins is not only dependent on diffusion to the interface, but the interfacial energy must be large enough to overcome an activation energy barrier for protein penetration and rearrangement into fluid–fluid interfaces. Thus, the surface forces, both of shear and dilation, may provide a means of altering protein interactions at the air–water interface [18,106,107] and, probably, at the oil–water interface as well. However, this effect was not observed in this study (at a frequency of 100 MHz and at an amplitude of compression/expansion cycle of 15%). The time dependence of interfacial pressure for heat-treated and native WPI was also studied [95]. The results obtained at the oil–water interface indicate a greater time dependence in the interfacial pressure and surface dilatational modulus. This may be associated with the fact that for heat-treated WPI, the level of protein unfolding is already maximal, although some protein aggregation could also be taking place [24,108]. The unfolding of the protein upon adsorption, especially for heat-treated protein, increases the accessibility of the sulfydryl group and the potential for formation of intermolecular disulfide cross-links which are responsible for the high E value of heat-treated adsorbed WPI films [95]. As observed by Das and Kinsella [109], the surface hydrophobicity—a measure of alteration of the native structure of a protein —of ␤ -lactoglobulin increases with heating at 80⬚C just as does the amount of protein adsorbed on emulsion droplets with the formation of multilayers. Dickinson and Hong [110] observed similar behavior for heat-treated ␤ -lactoglobulin at 70⬚C in relation to time-dependent surface shear viscosity and interfacial surface coverage in emulsion droplets. The effect of gelation on the viscoelastic characteristics of adsorbed protein films at the oil–water interface is of theoretical and practical importance and will be considered more fully later. B.

Protein Adsorption at the Air–Water Interface

Figure 11 shows that the adsorption kinetics of BSA from water and aqueous solutions of ethanol and sucrose at short adsorption time [99], up to approximately 2 s, are controlled by the diffusion of the protein toward the interface, since the surface pressure is lower than about 5 mN/m, in agreement with the Ward and Tordai model (Table 1). The presence of sucrose in the aqueous phase increased the rate of BSA diffusion toward the interface, but the opposite was observed for aqueous solutions of ethanol, especially at higher concentrations of this reagent in the bulk phase. The presence of ethanol in the bulk phase apparently introduces an energy barrier for the BSA diffusion toward the interface. This could be attributable to a competition with previously adsorbed ethanol molecules for the penetration of the protein into the interface. In addition, if ethanol causes denaturation and/or aggregation of the protein in the bulk phase [111], the diffusion of the protein toward the interface could be diminished. Thus, the causes of the higher rate of BSA diffusion from aqueous solutions of sucrose, in comparison with that observed for water, must be different in aqueous ethanol solutions. Since protein molecules are preferentially hydrated in the presence of sucrose [112,113], it is possible that sucrose limits protein unfolding in the bulk phase [18] and reduces protein–protein interactions in the bulk phase © 2003 by Marcel Dekker, Inc.


873

FIG. 11 Rate of BSA adsorption at the air–aqueous phase interface at 20⬚C and pH 7; I: 0.05 M. Protein concentration: 0.1%wt/wt. Aqueous phase composition: water (a), ethanol 1.0 M (b), and sucrose 0.5 M (c).

and at the interface [17,18,92]. Both of these phenomena may increase the rate of BSA diffusion toward the interface. At longer adsorption time, in the period after that affected by the diffusion, an energy barrier exists against the BSA adsorption. This could be attributed to the penetration, unfolding, and rearrangements of the protein at the interface [99]. Clearly, the kinetics of adsorption of proteins at interfaces are highly complex, especially in the presence of typical food solutes such as ethanol and sucrose in the aqueous phase. Simple unfolding at the interface with configurational rearrangements of adsorbed protein molecules produces changes in surface pressure which can be represented by a single exponential with time. However, when an energy barrier exists, the rate of protein penetration into the surface film will be rate limiting. Ward and Regan [114] have used a modified form of the Ward and Tordai equation to monitor this process: © 2003 by Marcel Dekker, Inc.

874


TABLE 1 Characteristic Parameters for the Diffusion of BSA at the Air–Aqueous Phase Interface at 20⬚C System BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA BSA

(1.10⫺3%)–water (7.5.10⫺3%)–water (5.10⫺3%)–water (1.10⫺2%)–water (7.5.10⫺2%)–water (5.10⫺2%)–water (1.10⫺1%)–water (7.5.10⫺1%)–water (5.10⫺1%)–water (1%)–water (1%)–ethanol 0.5 M (1%)–ethanol 1 M (1.10⫺1%)–ethanol 0.1 M (1.10⫺1%–ethanol 0.25 M (1.10⫺1%)–ethanol 0.5 M (1.10⫺2%)–ethanol 0.5 M (1.10⫺1%)–ethanol 0.75 M (1.10⫺1%)–ethanol 1 M (1.10⫺2%)–ethanol 1 M (1.10⫺1%)–ethanol 1.5 M (1.10⫺1%)–ethanol 2 M (1.10⫺1%)–sucrose 0.1 M (1.10⫺1%)–sucrose 0.25 M (1%)–sucrose 0.5 M (1.10⫺1%)–sucrose 0.5 M (1.10⫺2%)–sucrose 0.5 M (1.10⫺1%)–sucrose 0.75 M (1.10⫺1%)–sucrose 1 M (1.10⫺1%)–ethanol 1 M ⫹ sucrose 0.5 M

ln

冉冊 d␲ d␪

= ln(k⬘␯ Co) ⫺

␲⌬A KT

Slope ␲ versus ␪1/2 (mN⭈m⫺1⭈s⫺1/2

LR

0.066 0.55 0.53 0.64 1.23 0.97 1.89 7.83 5.38 10.2 9.92 8.90 1.42 1.38 1.36 0.44 1.28 0.92 0.34 0.74 0.66 2.16 2.24 10.9 2.36 0.89 2.41 2.37 1.38

0.907 0.956 0.971 0.897 0.942 0.928 0.922 0.943 0.961 0.935 0.921 0.983 0.947 0.909 0.948 0.965 0.960 0.961 0.917 0.926 0.939 0.977 0.952 0.912 0.940 0.936 0.964 0.911 0.909

(1)

where k⬘ is the rate constant of adsorption, K is the Boltzman constant, ⌬A is the molecular area required for the molecule to adsorb at the interface, and ␯ is the number of adsorbing groups per protein molecule. A plot of ln(d␲ /d␪ ) versus ␲ must be linear. The long-term adsorption of BSA at the air–water interface, as an example, is given in Fig. 12. Different temperatures, protein concentrations in the bulk phase, and aqueous phase composition (aqueous solutions of ethanol and sucrose) give similar results [92,93]. We find, for all experiments of BSA adsorption, two or more linear regions in the semi-log plot of ln[(␲⬁ ⫺ ␲␪)/(␲⬁ ⫺ ␲0)] versus ␪ or in the plot of ln(d␲ /d␪ ) versus ␲ [Eq. (1)]. © 2003 by Marcel Dekker, Inc.


875

FIG. 12 Rate of BSA adsorption at the air–water interface at 20⬚C. Protein concentration: 10⫺2%wt/wt. The subphase was phosphate buffer (pH 7; I: 0.05 M ).

To summarize the effect of temperature, protein concentration, and aqueous phase composition on the time dependence of surface pressure during BSA adsorption from the bulk phase, the following conclusions were drawn [11,92,93] from the ln[(␲⬁ ⫺ ␲␪)/(␲⬁ ⫺ ␲0)] versus ␪ plots. The rate of BSA adsorption at the interface increased with both BSA concentration in the aqueous phase and temperature. With ethanol in the subphase the existence of an induction period was observed, which could be associated with the competitive adsorption of BSA on an ethanol film— BSA has a higher affinity than ethanol for the interface due to its higher hydrophobicity. This phenomenon could also reflect the existence of BSA–solute interactions in the aqueous phase or at the interface. Sucrose has no affinity for the interface but instead a strong cohesive interaction with water molecules. Protein molecules are preferentially hydrated in the presence of sucrose, limiting the protein unfolding and protein–protein interactions and, consequently, the rate of BSA adsorption was observed to increase when sucrose was present in the bulk phase.

IV.

COMPARISON OF ADSORBED AND SPREAD PROTEIN FILMS AT EQUILIBRIUM

A.

Adsorbed Protein Films at the Oil–Water Interface

The effect of protein concentration on the equilibrium surface pressure is shown in Fig. 13 [28,72]. This adsorption isotherm was deduced from the data shown in Fig. 9 by extrapolating the plot of interfacial tension versus ␪ ⫺1/2 to 1/␪ 1/2 = 0 [115,116]. As expected, the surface pressure (Fig. 13) and the surface dilatational modulus [28] increased markedly as the amount of emulsifier in the bulk phase increased. For WPI adsorbed films this continued until the protein concentration in the bulk phase approached a critical value as indicated by the achievement of a plateau value with no © 2003 by Marcel Dekker, Inc.

876


FIG. 13 Adsorption isotherm for WPI at the oil–water interface at 20⬚C and pH 5; I: 0.05 M.

further change as the concentration was increased. The break point in the curve of surface pressure as a function of bulk concentration suggests the existence of a critical concentration, at which the interactions between WPI molecules lead to a constant surface activity until the interface is saturated with protein. The concentration of protein at the break point represents a critical concentration that determines the existence of protein–lipid interactions at the oil–water interface [72]. B.

Adsorbed and Spread Protein Films at the Air–Water Interface

Equilibrium spreading pressure of proteins (␤ -casein, caseinate, and WPI) at the air– water interface in the temperature range between 5 and 40⬚C is shown in Fig. 14 [28,117]. The magnitude of ␲e was dependent on the protein and the temperature. It can be seen that ␲e for ␤ -casein and caseinate was not affected by temperature over

FIG. 14 Temperature dependence of equilibrium surface pressure (␲e) for spread monolayers of ␤ -casein (䡩), caseinate (䉭), and WPI (䉮) on air–water.



877

the range plotted. However, ␲e for WPI increased slightly with temperature and more markedly at temperatures higher than 25⬚C. This phenomenon may be related to the temperature dependence on WPI spread films [105]. Figure 15 shows adsorption isotherms for proteins on water at 20⬚C [28,117]. The protein concentration dependence of surface pressure showed classical sigmoidal behavior. At low protein concentrations, the initial solutions caused only a small increase in the surface pressure. Thereafter, the surface pressure increased with protein concentration and then tended to a plateau. The beginning of this plateau occurred over the protein concentration range from 10⫺3 –1%wt/wt. Some differences existed between proteins. ␤ -Casein showed significant surface activity at protein

FIG. 15 Adsorption isotherm for ␤ -casein (a), caseinate (b), and WPI (c) on buffered water at 20⬚C and pH 7; I: 0.05 M. Different symbols are for repetitive experiments. The equilibrium spreading pressures for proteins are indicated by the arrows.


878


concentrations in the bulk phase of 10⫺6%, and surface pressure increased with protein concentration up to 1%wt/wt. On the other hand, the value of ␲e for spread film was lower than that for the surface pressure at the plateau for an adsorbed film. Caseinate showed significant surface activity at concentrations in the bulk phase of 10⫺5%wt/wt. This concentration was one order of magnitude higher than for ␤ casein. That is, at low concentrations, the surface activity of the proteins comprising caseinate gave lower surface activity than ␤ -casein alone. However, at the higher concentrations, the surface activity of the proteins forming caseinate gave higher surface activity than ␤ -casein alone, behavior similar to that observed with spread films (Fig. 14). These results suggest that the individual casein components in caseinate adsorb independently to the air–water interface, with few interactions between them. Finally, WPI showed significant surface activity at concentrations in the bulk phase of 5 ⫻ 10⫺5%wt/wt. The surface pressure increased with protein concentration and was tending to a plateau at the maximum protein concentration employed in the bulk phase (5%wt/wt). The value of ␲e for spread WPI film was lower than that of the surface pressure of adsorbed film at the maximum protein concentration in the bulk phase (5%wt/wt). The behavior of the adsorbed protein films (Fig. 15) can be interpreted in terms of monolayer coverage (see also Fig. 7). At the lower protein concentrations, as the surface pressure is close to zero, the adsorbed protein residues may be considered as a two-dimensional ideal gas. Proteins at higher concentrations, but lower than that of the plateau, form a monolayer of irreversibly adsorbed molecules. As the plateau is attained, the monolayer is saturated by protein that is irreversibly adsorbed. At higher protein concentrations, the protein molecules form multilayers beneath the primary monolayer, but these structures do not contribute significantly to surface pressure [75]. The presence of multilayers at the maximum protein concentration in the bulk phase has been deduced from Brewster angle microscopy (Fig. 8) and surface dilatational rheology (Fig. 17). Finally, differences observed between surface pressure at the plateau for adsorbed protein and equilibrium spreading pressure should be associated with a different rearrangement of residues when the protein is either adsorbed or spread on the interface at the highest surface density [105].

V.

RELAXATION AND VISCOELASTIC BEHAVIOR

A.

Long-Term Relaxation Phenomena of Spread Protein Films at the Air–Water Interface

Nonequilibrium processes occurring in systems containing fluid–fluid interfaces with a surfactant present are of great practical significance and include important technological operations such as emulsification and foaming. Two experimental approaches can be used for the analysis of long-term relaxation phenomena in emulsifier monolayers. In the first, the surface pressure (␲ ) is kept constant, and the area (A) is measured as a function of time. This relaxation experiment is the usual, preferred method, and is capable of being interpreted kinetically [118]. In the second approach, area is kept constant (at the collapse), and the decrease in surface pressure is monitored as a function of time. Information on the various relaxation paths detailed in Fig. 16 can be derived from the data. © 2003 by Marcel Dekker, Inc.


879

FIG. 16 Relaxation mechanisms in spread monolayers at the air–water interface, where ␲ and ␲e are the surface pressure and the equilibrium surface pressure, respectively.

Protein monolayers behave differently from typical lipids under the same experimental conditions. For protein monolayers an attempt was made to interpret the relaxation as a first-order monolayer molecular loss or collapse (Fig. 16). However, whether obtained from area measurement or surface pressure decays, fits of the experimental data at surfaces pressures lower than and higher than the equilibrium surface pressure (␲e) require two exponential decays [28,119,120]. The relaxation of protein monolayer therefore is not a simple process. At surface pressures lower than ␲e, the relaxation rate and the amplitude of the area relaxation depend on the surface pressure. The relaxation rate (quantified by means of the relaxation time, ␶, inverse rate constant) is higher at the highest surface pressure. The amplitude of the area relaxation increased with surface pressure. This is a reversible process. In fact, at surface pressures lower than ␲e no relaxation phenomena were observed in the ␲-A isotherm [119,120]. Nor was there any hysteresis during continuous compression–expansion cycles in this region. Protein monolayer stability was also tested under the most adverse conditions, at the maximum interfacial density (at constant collapse area). Under these conditions, the relaxation phenomena in protein monolayers are controlled predominantly by the collapse mechanism [119,120]. At an increased relaxation time, the surface pressure tends to a plateau that is practically coincident with the value of ␲e. As suggested by Graham and Phillips [75], the formation of multilayers of protein molecules under collapse conditions is more likely. In summary, the relaxation in relative molecular area at ␲ < ␲e and in surface pressure at ␲ > ␲e —which is mainly limited to the first 50 min—should be attributed to processes related to monolayer organization/reordering and collapse, respectively (see Fig. 16). © 2003 by Marcel Dekker, Inc.

880

B.


Surface Dilatational Characteristics of Spread Protein Films at the Air–Water Interface

There are many experimental devices for measuring dilatational rheology [80,91,121–125]. Convenient techniques for measuring surface dilatational rheology are derived from the longitudinal wave method [126]. Trough methods are preferable for insoluble films at the air–water interface. In these methods the area change applied to the film may be oscillatory, a step change, or a continuous expansion– compression. The mechanically generated longitudinal or capillary wave—such as that obtained for the movement of the barriers in a Langmuir trough containing the film—produces a response of surface tension which is monitored by a probe (i.e., a Wilhelmy plate) some distance away from the barrier [127,128]. To obtain surface rheological parameters—such as surface dilatational modulus (E ), elastic (Ed) and viscous (Ev) components, and loss angle tangent—a modified Wilhelmy-type film balance (KSV 3000) has been employed [69,129]. In this method the surface is subjected to small periodic sinusoidal compressions and expansions by means of two oscillating barriers at a given frequency (␻) and amplitude (⌬A/ A), and the response of the surface pressure is monitored (␲ ). Surface pressure was directly measured by means of two roughened platinum plates that can be situated anywhere on the surface between the two barriers. With this device, an isotropic dilatational deformation of the surface, without interference of shear, can be achieved, as demonstrated by the results derived for monoglyceride spread monolayers at the air–water interface [129]. In fact, the sinusoidal response in surface pressure due to sinusoidal area deformation was both the same in the two barriers and independent of the position of the barriers along the length of the film balance. The surface dilatational modulus for WPI monolayers (Fig. 17) increased with increasing surface pressure up to the collapse point [69]. This increase is a result of an increase in the interactions between the monolayer molecules, as deduced from ␲-A isotherms (Fig. 6), BAM images [72], and monolayer thickness (Fig. 8). How-

FIG. 17 Surface dilatational modulus as a function of surface pressure for ␤ -casein (䡩), caseinate (䉭), and WPI spread films (䉮) at the air–water interface at 20⬚C and pH 5; I: 0.05 M.



881

ever, for the more disordered protein (␤ -casein and caseinate) the E-␲ dependence is more complex. E increased to a maximum with surface pressure for structure 1, but decreased with surface pressure and passed to a minimum with structure 2. Finally, E increased up to the collapse point. The same inflexion in the ␲-⌫ curve was observed in the range of 10–20 mN/m and was attributed to the transition from an all-train configuration to a train-and-loop conformation of the ␤ -casein molecule [130]. The results with protein monolayers indicate that the dilatational modulus is not only determined by the structure of protein molecules, but the internal nature of the spread protein molecule also plays an important role. In fact, for the more ordered ␤ -lactoglobulin molecules in WPI the surface dilatational modulus is higher than that for ␤ -casein or caseinate molecules (Fig. 17) with disordered structure at the same surface pressures. Changes in surface dilatational properties for ␤ -casein monolayer (as an example) as a function of frequency of oscillation over a range of 1 to 300 mHz, at a representative surface pressure (20 mN/m), are illustrated in Fig. 18 [69]. It can be seen that (1) the dilatational modulus increased with the frequency, (2) the dilatational modulus and its elastic component are essentially the same at frequencies lower than 50 mHz. However, significant differences between both rheological parameters were observed at frequencies higher than 50 mHz, mainly due to the decrease of the elastic component at increasing frequencies. (3) The value of the viscous component increased with the frequency and exceeded that of the elastic component at higher frequencies (␻ > 200 mHz). From these results it can be concluded that ␤ -casein monolayers present rheological behavior in dilatational conditions that is essentially elastic at low frequencies (␻ < 50) and viscoelastic at higher frequencies (␻ > 50). As a consequence of the viscoelastic behavior, the tangent of the loss angle increased with frequency (Fig. 18). This behavior was observed with caseinate and WPI [69] and with BSA under dilatational deformation in a ring trough [131]. The

FIG. 18 Frequency dependence of surface rheological parameters—surface dilatational modulus, E (䡩); surface dilatational elasticity, Ed (䉭); surface dilatational viscosity, EV (䉮); and loss angle tangent (〫)—for ␤ -casein spread films at the air–water interface. Surface pressure: 20 mN/m; temperature: 20⬚C; pH 5; I: 0.05 M; amplitude 5%.


882


frequency dependence of E was analyzed with the help of data derived from longterm relaxation phenomena discussed in the previous section [69].

VI.

GELATION OF ADSORBED PROTEIN FILMS AT THE OIL–WATER INTERFACE

The stability and physicochemical properties of oil-in-water emulsions stabilized by WPI—protein surface coverage, surface shear viscosity, and stability [110,132–134] —are particularly sensitive to their thermal history [132]. Even though the phenomenon is not well understood [135,136], the interfacial gelation of globular proteins may possess great technological importance, because many protein-stabilized emulsions undergo some degree of thermal exposure during their processing, storage, or usage. We have therefore analyzed the viscoelastic characteristics of WPI heat-induced gels at the oil–water interface as a function of temperature and heating conditions [105,137]. Heat-induced interfacial aggregation of WPI, previously adsorbed at the oil– water interface, was studied by measuring interfacial dynamic characteristics (interfacial tension and surface dilatational properties) in an automatic drop tensiometer [105,137]. Dynamic rheological measurements in which sinusoidal oscillating stress or strain is applied to the sample are the preferred methods for characterizing viscoelastic foods [138,139]. These dynamic measurements allow coagulation and gelation to be monitored since the induced deformations are usually so small that their effect on structure is negligible. Figure 19 exemplifies the time-dependent interfacial tension and surface dilational properties of WPI adsorbed films on the oil–water interface as a consequence of thermal treatment for concentrations of protein in the bulk phase of 10⫺1%wt/wt. Overall, WPI adsorbed films behaved qualitatively in a similar manner after similar heat treatment, no matter what the protein concentration in the bulk phase, over the range 10⫺1 and 10⫺5% wt/wt. Briefly, E decreased during heating, passed through a minimum, and then increased as the heating progressed and tended to a plateau value just at the end of the heating period before the isothermal treatment. During the isothermal treatment (at 40, 60, and 80⬚C), E tended to increase to a plateau, especially during the first period at 40⬚C. The E value at the plateau decreased after thermal treatment at 60 and 80⬚C due to the effect of temperature on rheological parameters [137]. The surface activity was increased with the heat treatment because the conformational changes of molecules during heating may include further unfolding, reorganization, and aggregation of the molecules to bring more hydrophobic segments from the interior of the molecule to the oil–water interface. The rate of thermally induced changes in WPI adsorbed films at the oil–water interface increased with protein concentration in solution. Data in Fig. 19, especially the time dependence of E and phase angle, can be used in a quantitative kinetic analysis of the gelation process at the interface. If the interfacial gelation follows consecutive first-order steps, as for protein in solution [140], a first-order kinetics equation can be used to monitor the time dependence of E during heating. By plotting the data as ln E versus t it is possible to identify two linear regression regions [105], which account for the aggregation and/or cross-linking steps occurring consecutively and/or concurrently. The rate constants increased significantly as the protein concen© 2003 by Marcel Dekker, Inc.


883

FIG. 19 (a) Time evolution of temperature (——), interfacial tension (*), and (b) surface rheological properties—surface dilational modulus, E (䡩) and phase angle (▫)—for WPI adsorbed film during the isothermal treatment at 80⬚C. Protein concentration in solution: 10⫺1%; pH 5; I: 0.05 M; frequency of oscillation: 100 mHz; amplitude of sinusoidal oscillation: 15%.

tration in solution increased. That is, conformational changes and protein aggregation increase with protein–protein interactions, which increase with protein concentration in solution. Thus it can be concluded that the gelling time, quantified by the rate constants for E–time dependence, is a useful parameter for detecting transition points in heat-treated protein films at the oil–water interface, just as for gelation in solution [141,142]. An important characteristic of WPI interfacial gelation, both from a theoretical and practical point of view, is the low protein concentration in solution necessary for interfacial gelation in comparison with that necessary for WPI gelation in solution. In this study we have observed the existence of significant changes in interfacial dynamic properties associated with WPI gelation in adsorbed films at the oil–water interface at protein concentration in solution as low as 10⫺5%wt/wt, even at temperatures of 40⬚C. These values are far below those between 1 and 2.5% required for © 2003 by Marcel Dekker, Inc.

884


gelation in solution under similar conditions of pH [138,139,143,144], but it must be remembered that the local protein concentration at the interface is far above its concentration in solution. As observed in the experimental data (Fig. 6) the interfacial tension decreases steadily with heat treatment, which means that the interfacial activity of WPI films increases in the same way. At each protein concentration in solution, the minimum interfacial tension coincided with the maximum E value. The trend observed in interfacial characteristics (interfacial tension and viscoelastic characteristics) of WPI adsorbed films with the heat treatment is similar to the influence of temperature on the interactions and stability of WPI stabilized oil-in-water emulsions [65,132,134,145,146].

VII. A.

INTERACTIONS AND STABILITY Interactions Between Protein-Coated Surfaces

The scf approach described earlier has also been used to estimate the interaction energy between a pair of parallel hydrophobic surfaces in the presence of adsorbing ␤ -casein or ␣S1-casein [53]. Protein, solvent, and small ions were introduced into the gap between parallel hydrophobic surfaces and the system allowed to reach full thermodynamic equilibrium. Closing the gap and permitting equilibrium adsorption to be attained allows the interaction between the coated surfaces to be calculated as the difference from that calculated at infinite separation. It was found that the calculated interaction energy remained positive and repulsive for all separations for ␤ -casein irrespective of pH or ionic strength. Direct experimental measurements of interaction forces between ␤ -caseincoated surfaces are obtainable as a function of surface separation by using the interferometric surface force apparatus [147]. On hydrophobized mica, a long-range repulsive electrostatic force was observed between ␤ -casein layers. This could be represented by a DLVO-type potential with the plane of charge and origin of the van der Waals forces placed at the onset of the steric wall determined as 4 nm thick. As the surface separation was decreased, this long-range repulsion was overcome by an attractive force at a surface separation of about 25 nm. This caused the proteincovered surfaces to slide into contact, the force required to separate the adhering surfaces exceeding 2.4 mN⭈ m⫺1. The scf calculations show that reducing pH toward the protein isoelectric point reduces the strength and range of the interlayer repulsion. This implies a substantial electrostatic contribution to the calculated interaction energy, but the energy remains positive at all separations. There is no attraction corresponding to that seen with the surface force apparatus. The origin of this discrepancy remains obscure. On the one hand, the calculations report the result of an equilibrium being achieved on adsorbing to hydrophobic surfaces at the separation of interest. This is not conceptually identical to preadsorbing the protein at infinite separation and then closing the gap. Nor does the computed interaction include any contribution from ubiquitous van der Waals forces. It only includes that part of the overall interaction due to the casein chains. On the other hand, Nylander and Wahlgren [147] also performed measurements with ␤ -casein preadsorbed onto hydrophilic surfaces. On bringing these surfaces together, the inward jump was no longer evident. The force was entirely repulsive and the same force was observed on decompression as on compression. The © 2003 by Marcel Dekker, Inc.


885

exponential decay again suggested double-layer forces contributed to this repulsive interaction. When the ␤ -casein solution was replaced with a 1 mM NaCl buffer, a substantial change was observed in the force–distance profile. Not only was the magnitude of the double-layer force reduced, but also an attractive force became apparent at surface separations below 30 nm. Also the force curves measured for ␤casein adsorbed onto the hydrophobic surface could be superimposed on that obtained with the hydrophilic counterpart provided it was shifted 4.7 nm outward. Nylander and Wahlgren [147] suggested that ␤ -casein forms a bilayer on the hydrophilic mica; the first arrivals adsorbing flat onto the surface, thus exposing hydrophobic surfaces outward. This allows molecules forming the second layer to act as if they were adsorbing onto a hydrophobic surface and to orient in such a way that their hydrophilic parts protrude into solution. The final result is that, from the perspective of the bulk solution, the exposed interfaces look similar on the two surfaces.

B.

Comparison of ␣S1-Casein and ␤ -Casein Behavior

The scf calculations predict differences in interaction behavior for the two proteins ␤ -casein and ␣S1-casein. This main difference in the predicted interaction behavior occurs at higher ionic strengths. While the potential energy remains repulsive for ␤ -casein at all separations, the theory predicts for ␣S1-casein a potential well at pH 5.5 which deepens as ionic strength is increased in the range 50–200 mM. The underlying physical mechanism thought to be the origin of this attraction is the possible bridging of the gap by the ␣S1-casein molecules with their two sticky ends. This bridging effect computes into the interaction as an attractive contribution, entropic in origin in thermodynamic terms. This predicted interaction behavior for the two proteins is in broad qualitative agreement with what is actually observed in model oil–water emulsions [148–150]. Those prepared with ␤ -casein as sole emulsifier are stable toward NaCl addition (>2 M ), whereas those prepared with ␣S1casein become extensively flocculated in 0.1–0.2 M NaCl. Casanova and Dickinson [149] have also shown that ␤ -casein in mixed model films of ␣S1-casein and ␤ casein has a strongly protective effect in overcoming the tendency of ␣S1-casein emulsions to flocculate at these low ionic strengths. While this agreement is gratifying, it may be illusory. We still have no reliable information on the form of the total interaction potential from which we might hope to quantitatively predict stability behavior. In keeping with the spirit of colloidal stability theory, we would imagine this interaction potential to be a balance of repulsive and attractive forces, perhaps as depicted in Fig. 20. It would be naive, however, to present this as a simple summation of van der Waals attraction, steric, and electrostatic repulsion terms. According to Nylander and Wahlgren [147], the long-range repulsion is electrostatic in origin, but in overcoming this barrier they entered an attractive well of unknown depth. The dynamic light scattering data of Brooksbank et al. [42] showed the steric stabilizing layer to be thinned before it was overcome and aggregation ensued. The position as well as the height of the steric stabilizing barrier are thus under electrostatic control. This coupling of steric and electrostatic influences lies at the heart of the representational problem and remains a challenge for the theoreticians. © 2003 by Marcel Dekker, Inc.

886


FIG. 20 Illustration of possible form of interaction potential function with steric barrier to be overcome before entry into an attractive potential well. Position as well as height of barrier could be functions of electrostatic repulsion, with barrier moving inward as electrostatic repulsion was shielded or neutralized. Shifts could be largely independent of barrier height (or at least couple with height), suggesting that simply additive steric and electrostatic terms may not be appropriate for prediction of an overall potential energy function.

C.

Interactions Between ␤ -Lactoglobulin-Coated Surfaces

Because the interfacial films are relatively thin, it is unlikely that ␤ -lactoglobulin layers stabilize emulsions by the same steric stabilization mechanism invoked for ␤casein. Ismailova, Yamployskaya, and Tulovskaya [151] revived Rehbinder’s concept and related coalescence stability directly to the mechanical properties of the adsorbed protein layer. The Rehbinder hypothesis was developed further, showing that it was not simply the magnitude of the surface elastic modulus that was important, but rather the stress at which the film was ruptured. It is also likely that electrostatic interactions play a part in ensuring the stability of ␤ -lactoglobulin-stabilized emulsions, since such emulsion droplets are highly susceptible to flocculation by calcium ions even though the native whey protein is not itself precipitated by ionic calcium in aqueous solution [152–154].

VIII.

COMPETITIVE DISPLACEMENT

Milk proteins saturate the fluid interface at lower concentrations than small molecule surfactants [155], and hence the protein component dominates in mixture situations where the surfactant concentration is low. On the other hand, at high surfactant concentrations, more efficient packing of the surfactant molecules leads to lower interfacial tension from the surfactants, so the protein is displaced from the interface. The detailed structure and composition of the mixed layer depends on the balance © 2003 by Marcel Dekker, Inc.


887

of all interactions between the components both at the interface and in the bulk solution [14]. Nonionic water-soluble surfactants are generally more effective than oil-soluble surfactants at displacing proteins from the oil–water interface. For an anionic surfactant like sodium dodecyl sulfate, a much higher concentration of surfactant is required for displacement because of its propensity for complex formation with the protein both in solution and at the interface. All of these competitive adsorption effects which can occur in real systems have implications for the stability and shelf-life of milk protein–based emulsion products. A.

Chemistry

Protein displacement is readily quantified by determining the protein content of the serum phase of the emulsion in the presence of added surfactant. Hydrocarbon oils are frequently used in model emulsion systems, and there is some evidence to suggest that the hydrophobic phase has little influence on the conformational structure that the casein adopts at the interface [57]. The susceptibility of the proteins to displacement by surfactant is, however, determined by the oil phase. For example, Leaver et al. [156] found that Tween 20 readily displaced ␤ -casein from fresh tetradecane emulsions but was less efficient from soya oil. Moreover, the displacement ability declined as the ␤-casein soya oil emulsions aged or following a heating regime of 80⬚C for 30 min, whereas neither aging nor heating had any effect on ␤ -caseinstabilized tetradecane emulsions. Susceptibility to displacement was also found to be pH dependent and a function of the casein type involved [157]. Reversed phase highpressure liquid chromatography analysis showed significant changes in peak shape and position in those proteins displaced from aged soya oil emulsions and relatively unchanged peaks from tetradecane emulsions over the same time scale. Tryptic digestion of the displaced proteins also showed modification of the primary structure, which increased with emulsion age. Gas chromatography–mass spectrometry of steam distillates from the emulsions showed the presence of a variety of aldehydes in soya oil emulsions that were not present in the original oil or in the tetradecane emulsions [158]. Several of these volatile aldehydes were identified as enals (␣, ␤ unsaturated aldehydes) which react readily with lysine side chains, providing them with a hydrophobic tail which assists in anchoring the protein into the oil phase, thereby decreasing its susceptibility to displacement. B.

Physical Methods

Neutron reflectivity has also been used to study competitive displacement of ␤ lactoglobulin and ␤ -casein from the air–water interface by the nonionic surfactant, hexaoxyethylene n-dodecyl ether (C12E6). The experiments for ␤ -lactoglobulin took advantage of the differing neutron scattering cross-sections for hydrogen and deuterium in hydrogenated and partially deuterated surfactant to determine the amounts of surfactant and protein adsorbed and, importantly, the location of the surfactant [159]. The data are shown in Fig. 21 as a function of molar surfactant to protein ratio, R. In the presence of deuterated surfactant, we see the total amount of reflective material at the interface (i.e., protein ⫹ surfactant). In the presence of hydrogenated surfactant, we see only the protein displacement profile. Fitting the profiles with error functions, complementary for the protein and regular for the surfactant, we can © 2003 by Marcel Dekker, Inc.

888


FIG. 21 Behavior of parameters derived from neutron reflectivity measurements during displacement of ␤ -lactoglobulin by surfactant C12E6. Plots show adsorbed layer thickness and adsorbed amounts, derived from Guinier plots, as function of surfactant/protein molar ratio. Open symbols: with hydrogenated C12E6 ; closed symbols: with partially deuterated C12E6 . Solid line through solid squares is spline fit to guide the eye. Solid line through open inverted triangles is error function fit to protein displacement curve. Long dashed line is error function fit for adsorbed surfactant. Summed with displacement error function (short dashed line through solid triangles), it gives data points for total adsorbed amount in presence of deuterated (reflecting) surfactant.

reproduce the total adsorbed amount as their summation. An interesting feature here is that at R ⬇ 0.3, when the amount of surfactant has almost saturated, there is apparently still a substantial amount of ␤ -lactoglobulin at the interface. The rms thickness of the adsorbed films, calculated by Guinier analysis of the reflectivity data, is also shown in Fig. 21. This shows an increase in the thickness of the layer, both in the presence of deuterated and hydrogenated surfactant at a point at which surfactant adsorption had almost reached its plateau value, but at which it was estimated that some 60% of the original protein was still in the interfacial film. It was suggested that to produce these peaks in the layer thickness plots, the surfactant had formed an inner layer adjacent to the air–water interface, pushing off the protein. This location was deduced from the similar increase in thickness seen in the presence of the hydrogenated surfactant. Had this ‘‘invisible’’ surfactant layered on top of the protein, there would have been no measurable increase in thickness in its presence. On the high surfactant concentration side of the peak, the rms thickness becomes that for the deuterated surfactant layer alone. The displacement behavior of ␤ -lactoglobulin by nonionic surfactants has recently been visualized using atomic force microscopy at both air–water and oil– water interfaces [78,160]. These studies clearly show a demixing phenomenon and have led to the proposal of an ‘‘orogenic model’’ for protein displacement, where © 2003 by Marcel Dekker, Inc.


889

the surfactant domains exert a lateral surface pressure which compresses, buckles, and collapses the ␤ -lactoglobulin protein regions prior to their displacement. They thus complement the neutron reflectivity measurements described above. Further support comes from a computer simulation study by Wijmans and Dickinson [161], who used a Brownian dynamics approach to model the competitive displacement process. Their model takes a bonded monolayer and introduces more strongly adsorbing, nonbonded particles. Initially the displacer particles fill up the gaps in the original bonded monolayer network. As the surface concentration of the displacer particles increases, the network ‘‘holes’’ grow, the strands become thinner, and some network particles are pushed away from the surface. Part of the bonded network remains pinned to the surface, but large parts of the film buckle out into the bulk solution, forming a relatively thick layer. Eventually the surface becomes saturated with displacer particles and the protein network detaches into the solution. The simulations emphasize the importance of the rheology of the protein network to the process and highlight the role of repulsive forces between protein and surfactant in promoting the phase separation.

C.

Competitive Adsorption and Displacement in Protein–Lipid Mixed Systems

Competitive adsorption of proteins and lipids at fluid interfaces can affect the stability of food dispersions [2]. Thus knowledge of the lipids, proteins, and their mixtures at fluid–fluid interfaces is a key factor for the formation and stability of food dispersions (emulsions and foams). However, more important in some products is the effect of the small molecules in destabilizing the emulsion [21]. In the formulation of ice cream, the small molecule emulsifier is added to break the adsorbed layer of protein and allow the adsorption of fat to the surface of the air bubble. Thus, an important action of the small molecule emulsifiers is to promote the displacement of caseins from the interface. There have been many studies of protein–lipid interactions, particularly between proteins and soluble lipids, in relation to the formation and stability of food emulsions and foams, but much less is known about the details of protein-insoluble lipid interaction [14,29].

1.

Protein–Lipid Interactions at the Air–Water Interface at the Equilibrium Surface tension data have been reported for the milk protein systems (␤ -casein, caseinate, and WPI) and their mixtures with three food-permitted insoluble lipids (monopalmitin, monoolein, and monolaurin) at equilibrium [117]. These experiments mimicked the behavior of emulsifiers in food emulsions in which an oil-soluble lipid (monopalmitin, monoolein, or monolaurin) diffuses to the interface to interact with a protein film which has adsorbed from the aqueous bulk phase. The effect of the protein/lipid ratio on the surface activity of mixed protein–monoglyceride systems is shown in Fig. 22. In these experiments, monoglyceride spread on a previously adsorbed protein film was maintained constant. Therefore, the variation in protein/ lipid ratio is due to the protein added to the bulk phase in the range 5 to 1 ⫻ © 2003 by Marcel Dekker, Inc.

890


FIG. 22 Effect of the spreading of monopalmitin (䡩), monoolein (〫), and monolaurin (䉭) on a film of ␤ -casein (a), caseinate (b), and WPI (c) previously adsorbed on the air–water interface. Temperature 20⬚C. Lipid superficial density (molecule.nm⫺2): monopalmitin, 12; monoolein, 9.7; and monolaurin, 9.5. The arrows indicate the equilibrium spreading pressure, ␲e , for monopalmitin [␲e (MP)], monoolein [␲e (MO)], and monolaurin [␲e (ML)].

10⫺5%wt/wt. The monoglyceride density spread on the interface was higher than that required for the monolayer collapse, as was deduced from the ␲-A isotherm [162]. The protein concentration dependence on surface pressure for protein–monoglyceride mixed systems showed a sigmoidal behavior. However, the surface activity of the mixed systems depended on the protein/monoglyceride ratio and the monoglyceride spread on the interface (Fig. 22). For all protein–monoglyceride mixed films, the surface pressure values ap© 2003 by Marcel Dekker, Inc.


891

proach those of pure protein films at higher relative protein concentrations in the mixed systems, as the protein saturated the monolayer. However, significant differences were observed at lower protein/monoglyceride ratios, especially for protein– monolaurin mixed films. In fact, for protein–monolaurin mixed films, the surface pressure followed the same dependence on protein concentration as pure protein. For protein–monopalmitin and protein–monoolein mixed films, the protein/ lipid ratio dependence on surface pressure was more complex. At lower relative protein concentrations, the surface pressure tended to the equilibrium surface pressure (␲e) of the pure monopalmitin or monoolein, which is indicated in Fig. 22 by the arrows. However, the level of surface pressure at the minimum protein/monoglyceride ratio depended on the interfacial composition. At the intermediate range of protein–monoglyceride concentrations, significant further reduction in the surface pressure was observed. The effect resulted in an inflection in the surface pressure curve in the intermediate region. The general behavior described earlier for protein–monopalmitin and protein– monoolein mixed films at the air–water interface (Fig. 22) is essentially the same as that observed with other monoglyceride–protein mixed films [15]. Thus it appears that at higher protein relative concentrations in protein–monoglyceride mixed films the protein determines the surface activity, but at lower relative protein concentrations in the mixture monoglyceride (monopalmitin or monoolein) determines the surface activity of protein–monoglyceride mixed films. As the monoglyceride (monopalmitin or monoolein) surface densities spread here, the protein is removed and a collapsed monopalmitin or monoolein film saturates the interface. In the intermediate region, the surface activity is determined by the existence of protein and monoglyceride at the interface. However, what is more difficult to establish is the degree of interactions between film-forming components in the mixed film. Protein–monolaurin mixed films behave differently. In fact, the protein, at every protein/monolaurin ratio, determined the surface activity of protein–monolaurin mixed films. This phenomenon may be associated with the instability of monolaurin monolayers at the air–water interface. In fact, from relaxation experiments we have observed [163] a significant monolaurin monolayer molecular loss, which is more pronounced as the surface pressure increases, with a maximum at the surface density utilized in this study (at the collapse point). These results support the hypothesis that a reduced level of interaction exists between monolaurin and protein at the air–water interface. The extent of removal of protein by a surfactant is influenced by factors affecting the binding strength of a protein to a surface. Thus the displacement of protein by surfactant has been found to decrease in conditions favoring conformational changes. The ease of displacement, however, is not only influenced by protein properties, but also by the type of surfactant [14] and the aqueous composition [15,16]. In particular, as we have demonstrated here, the distribution of monoglyceride and proteins at the air–water interface is influenced by the spreading of monoglyceride on a previously adsorbed protein, especially when protein concentration in the bulk phase is higher than that required for complete coverage, at the plateau region (Fig. 22). Thus the way in which proteins and monoglycerides are spread or adsorbed on the interface may have a role in determining the interfacial characteristics of the mixed film. © 2003 by Marcel Dekker, Inc.

892


2.

Protein–Lipid Interactions at the Air–Water Interface at Steady State From the ␲-A isotherms it is clear that some degree of interaction exists between monoglycerides and proteins in spread mixed films, which becomes more pronounced as the monolayer is compressed at the highest surface pressures [19,94,164,165]. Recent experiments (including measuring the ␲-A isotherm coupled with Brewster angle microscopy and relative thickness of spread monolayers) have shown for the first time [19,94,164,165] that in monoglyceride–protein mixed films, islands of protein and monoglyceride exist at the air–water interface on a microscopic level, but with few interactions between them, depending on the surface pressure (Fig. 23). At surface pressures lower than those for protein collapse a mixed monolayer of monoglyceride and protein may exist (Fig. 23a). However, at surface pressures higher than that for protein collapse, the mixed monolayers are essentially dominated by monoglyceride molecules. At higher surface pressures, collapsed pro-

FIG. 23 Visualization of monopalmitin–WPI mixed monolayers by Brewster angle microscopy at 20⬚C and pH 5.0; I: 0.05 M. (a) At surface pressures lower than that for WPI collapse (␲ ⬵ 31 mN/m) there exists coexistence between small circular liquid-condensed monopalmitin domains and a homogeneous phase of liquid-expanded monopalmitin and WPI domains. (b) At higher surface pressures, near and after the WPI collapse (25 < ␲ < 37 mN/m), the squeezing out of WPI by monopalmitin can be distinguished in a region with LC domains of monopalmitin (dark area) over a sublayer of collapsed protein (bright area). (c) At the lipid collapse, the monopalmitin domains were so closely packed that the monolayer morphology acquired a high homogeneity. (d) In the region of highest surface pressure different regions of collapsed WPI were observed on the interface. Temperature 20⬚C. The horizontal direction of the image corresponds to 630 ␮m, and the vertical direction corresponds to 470 ␮m.



893

tein residues may be displaced from the interface by monoglyceride molecules (Fig. 23b). However, monoglyceride molecules are unable to completely displace protein residues from the air–water interface even at the highest surface pressure, at the collapse pressure of the monoglyceride (Figs. 23c and d). Over the overall range of existence of the mixed film the monolayer presents some heterogeneity due to the fact that domains of monoglyceride and protein residues are present during the monolayer compression, giving interfacial regions with different relative film thickness. Interactions, miscibility, and displacement of proteins by monoglycerides from the air–water interface depend on the particular protein–monoglyceride system. Different proteins show different interfacial morphology, confirming the importance of protein secondary structure in determining the mechanism of interfacial interactions. On the other hand, the lower surface activity of monoolein justifies the idea that this lipid has a lower capacity than monopalmitin for protein displacement. In fact, monoolein requires higher surface pressures than monopalmitin for protein displacement from the air–water interface [165]. The prevalence of monoglyceride increases with its concentration in the mixture and at higher surface pressures. In summary, on a microscopic level, the distribution of lipids and proteins in mixed spread films at the air–water interface is heterogeneous and depends on the surface pressure and the lipid–protein ratio in the mixed film. Until recently the exact mechanism for protein displacement was unclear. Recent studies [78,160,166] have shown that the displacement of milk proteins from the air–water interface by water-soluble surfactants (Tween 20 and SDS) involves a novel orogenic mechanism. Such a model provides an explanation for the displacement of protein by surfactants, but differences exist between the behavior of adsorbable water-soluble surfactants and spread water-insoluble monoglycerides. The results suggest that for spread monoglycerides [19,94,164,165] the first stage of the orogenic mechanism, which occurs at surface pressures lower than the equilibrium surface pressure of the protein, involves a displacement front of monoglyceride domains instead of the adsorption of water-soluble surfactant molecules at defects in the protein network. The second stage, which occurs at surface pressures near to and above the equilibrium surface pressure of the protein, involves a buckling of the monolayer and reordering of the molecules as the protein film gets thicker in response to the decreasing surface coverage. Finally, at sufficiently high surface pressures the protein network begins to fail, freeing proteins which then desorb from the interface [78,160,166]. But, for spread monoglyceride monolayers, the protein displacement is not total even at the highest surface pressure, at the collapse point of the mixed film. The orogenic displacement mechanism is a consequence of the low level of interaction between proteins and monoglycerides at the air–water interface [19,94,164,165]. 3.

Protein–Lipid Interactions at the Oil–Water Interface Under Dynamic Conditions

The existence of WPI–monoglyceride (monopalmitin and monoolein) interactions at the interface has been shown in measurements of interfacial tension and surface dilatational properties [105]. Systematic experimental studies of surface dynamic properties, as a function of time and at long-term adsorption, for protein–monoglyceride mixed films at the oil–water interface were carried out in an automated © 2003 by Marcel Dekker, Inc.

894


drop tensiometer. The dynamic behavior of protein ⫹ monoglyceride mixed films was found to depend on the adsorption time, the lipid and the protein/lipid ratio, all in a rather complicated manner. The protein determined the interfacial characteristics of the mixed film since, at WPI ⱖ 10⫺2%wt/wt, the protein saturated the film (Fig. 13), no matter what the concentration of the monoglyceride. However, a competitive adsorption of the emulsifier (WPI and monoglycerides) does exist, as the concentration of protein in the bulk phase is far lower than that for interfacial saturation. This critical concentration determines the existence of protein–lipid interactions at the oil–water interface. That is, monopalmitin and, especially, monoolein in the oil bulk phase are unable to displace totally the adsorbed WPI film. In fact, the protein dominated the interfacial characteristics of WPI–monoolein mixed films throughout the whole range of protein/lipid ratios employed. In these experiments, the existence of protein–monoglyceride interactions at the air–water interface has been demonstrated by different complementary techniques, including tensiometry under static and dynamic conditions, Langmuir- and Wilhelmy-type surface film balances, and Brewster angle microscopy. Surface tension measurement is an easier complementary experimental technique for providing information regarding the interfacial characteristics of pure protein and lipid films and the existence of protein–lipid interactions at the interface. However, surface pressure–concentration experiments are not sufficient to allow a full picture of the nature of protein–monoglyceride interactions at the interface. From the results derived from the different techniques, it can be concluded that the interfacial characteristics of mixed emulsifiers at fluid interfaces depend on the nature of the interface (either air–water or oil–water) and, in a complex fashion, on the way by which these emulsifiers are adsorbed to the interface (either by cooperative or competitive adsorption/spreading of the film-forming components). On the other hand, lowering of the surface (interfacial) tension by emulsifiers (proteins and lipids) is only a first step in the preparation of stable food emulsions and foams. A low surface (interfacial) tension facilitates breaking up the oil phase into smaller droplets. However, dispersion requires rapid and substantial stretching of bubbles or drops, and consequently the surface (interfacial) tension may be far from equilibrium. Thus, dynamic properties of adsorbed emulsifier layers are also important due to their stabilizing function during emulsification.

IX.

CONCLUDING REMARKS

In this chapter, we have analyzed the structure, morphology, adsorption, interactions, and dynamic properties of food dairy proteins (␤-casein, caseinate, and WPI) at the air–water and oil–water interfaces. The summary includes an assessment of information derived from a variety of chemical and physical techniques. Combined surface chemistry (Langmuir- and Wilhelmy-type film balances and dynamic tensiometry) and microscopy (Brewster angle microscopy, BAM) techniques have been used to analyze the surface characteristics of proteins, with structural information included from neutron reflectivity, dynamic light scattering, and enzyme accessibility measurements. The results demonstrate that protein type and temperature affect the interfacial characteristics. The nature of emulsifier interactions at the interface has an important role on their physicochemical characteristics, including their role in con© 2003 by Marcel Dekker, Inc.


895

ferring stability on emulsions and foams. Important functional differences have been demonstrated between globular (WPI) and disordered (␤ -casein and caseinate) proteins. The efficacy of the caseins as steric stabilizing agents of emulsions has been considered and the current state of theories regarding those interactions has been summarized. ACKNOWLEDGMENTS DSH acknowledges the support of the Scottish Executive Environment and Rural Affairs Department and BBSRC (UK) in this work. The research was also supported by the European Community through Grant FAIR-CT96-1216 (DSH, JMRP), by CICYT through Grant ALI97-1274-CE, and by DGICYT through Grant PB97-0734 (both JMRP). REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

15. 16. 17. 18. 19.

E. Dickinson, An Introduction to Food Colloids. Oxford: Oxford University Press, 1992. E. Dickinson. J. Dairy Sci. 80:2607–2619, 1997. E. Dickinson. J. Chem. Soc. Faraday Trans. 94:1657–1669, 1998. E. Dickinson. In Modern Aspects of Emulsion Science (B. P. Binks, ed.). Cambridge: Royal Society of Chemistry, 1998, pp. 145–174. E. Dickinson. Coll. Surf. B: Biointerfaces 15:161–176, 1999. E. Dickinson. Coll. Surf. B: Biointerfaces 20:197–210, 2001. S. Damodaran and A. Paraf. Food Proteins and their Applications, New York: Marcel Dekker, 1997. S. E. Friberg and K. Larsson, Food Emulsions, 3rd ed. New York: Marcel Dekker, 1997. R. Hartel and G. R. Hasenhuette, Food Emulsifiers and Their Applications. New York: Chapman and Hall, 1997. F. D. Gunstone and F. B. Padley, Lipid Technologies and Applications. New York: Marcel Dekker, 1997. R. E. McDonald and D. B. Min, Food Lipids and Health. New York: Marcel Dekker, 1996. ˜ Colloids Surf. B 15:235–252, 1999. J. M. Rodrı´guez Patino and M. R. Rodrı´guez Nino. J. Sjo¨blom, Emulsions and Emulsion Stability. New York: Marcel Dekker, 1996. M. Bos, T. Nylander, T. Arnebrandt, and D. C. Clark. In Food Emulsions and Their Applications (G. L. Hasenhuette and R. W. Hartel, eds.). New York: Chapman & Hall, 1997, pp. 95–146. ˜ and J. M. Rodrı´guez Patino. J. Am. Oil Chem. Soc. 75:1233– M. R. Rodrı´guez Nino 1240, 1998. ˜ and J. M. Rodrı´guez Patino. J. Am. Oil Chem. Soc. 75:1241– M. R. Rodrı´guez Nino 1248, 1998. ˜ , P. J. Wilde, D. C. Clark, F. A. Husband, and J. M. Rodrı´guez M. R. Rodrı´guez Nino Patino. J. Agric. Food Chem. 45:3010–3015, 1997. ˜ , P. J. Wilde, D. C. Clark, and J. M. Rodrı´guez Patino. J. Agric. M. R. Rodrı´guez Nino Food Chem. 45:3016–3021, 1997. ˜ J. Agric. Food Chem. J. M. Rodı´guez Patino, S. C. Carrera, and M. R. Rodrı´guez Nino. 47:4998–5008, 1999.


896


20.

˜ P. J. Wilde, D. C. Clark, and J. M. Rodrı´guez Patino. J. Agric. M. R. Rodrı´guez Nino, Food Chem. 46:2177–2184, 1998. ˜ P. J. Wilde, D. C. Clark, and J. M. Rodrı´guez Patino. Langmuir M. R. Rodrı´guez Nino, 14:2160–2166, 1998. K. G. Berger. In Food Emulsions (S. E. Friberg and K. Larsson, eds.). New York: Marcel Dekker, 1997, pp. 423–490. H. D. Goff and W. K. Jordan. J. Dairy Sci. 72:18–29, 1989. P. Cayot and D. Lorient. In Food Proteins and Their Applications (S. Damodaran, A. Paraf, eds.). New York: Marcel Dekker, 1997, pp. 225–256. D. G. Dalgleish. In Emulsions and Emulsion Stability (J. Sjo¨blom, ed.). New York: Marcel Dekker, 1996, pp. 287–325. D. G. Dalgleish. In Food Proteins and Their Applications (S. Damodaran and A. Paraf, eds.). New York: Marcel Dekker, 1997, pp. 199–223. T. Nylander and B. Ericsson. In Food Emulsions (S. E. Friberg and K. Larsson, eds.). New York: Marcel Dekker, 1997, pp. 189–233. ˜ S. C. Carrera, J. M. Rodrı´guez Patino, F. M. Cejudo, and N. M. R. Rodrı´guez Nino, J. M. Garcıá. Chem. Eng. Commun. 2000 (in press). R. Miller, V. F. Fainerman, A. V. Makievski, J. Kraˆgel, D. O. Grigoriev, V. K. Kazakov, and O. V. Sinyachenko. Adv. Colloid Interface Sci. 86:39–82, 2000. D. Mobius and R. Miller, Studies of Interface Science, Vol 7. Amsterdam: Elsevier, 2001. M. Malmsten, Biopolymers at Interfaces. New York: Marcel Dekker, 1998. D. S. Horne and J. Leaver. Food Hydrocolloids 9:91–95, 1995. D. S. Horne. In Encyclopaedia of Dairy Sciences (H. Singh, ed.). New York: Academic Press (in press). D. S. Horne. Int. Dairy J. 8:171–177, 1998. P. Walstra. In Gums and Stabilizers for the Food Industry, Vol. 4 (G. O. Phillips, D. J. Wedlock, and P. A. Williams, eds.). Oxford: IRL Press, 1988, pp. 323–336. J. Benjamins, J. A. DeFeijter, M. J. A. Evans, D. E. Graham, and M. C. Phillips. Faraday Discuss. Chem. Soc. 59:218–229, 1975. B. W. Morrissey and C. C. Han. J. Colloid Interface Sci. 65:423–431, 1978. T. Nylander and M. Wahlgren. J. Colloid Interface Sci. 162:151–162, 1994. T. Kull, T. Nylander, F. Tiberg, and N. M. Wahlgren. Langmuir 13:5141–5147, 1997. A. R. Mackie, J. Mingins, and A. N. North. J. Chem. Soc. Faraday Trans. 87:3043– 3049, 1991. D. G. Dalgleish. Colloids Surf. 46:141–155, 1990. D. V. Brooksbank, C. M. Davidson, D. S. Horne, and L. Leaver. J. Chem. Soc. Faraday Trans. 89:3419–3425, 1993. J. Penfold, R. M. Richardson, A. Zarbakhsh, J. R. P. Webster, D. G. Bucknall, A. R. Rennie, R. A. L. Jones, T. Cosgrove, R. K. Thomas, J. S. Higgins, P. D. I. Fletcher, E. Dickinson, S. J. Roser, I. A. McLure, A. R. Hillman, R. W. Richards, E. J. Staples, A. N. Burgess, E. A. Simister, and J. W. White. J. Chem. Soc. Faraday Trans. 93: 3899–3917, 1997. E. Dickinson, D. S. Horne, J. S. Phipps, and R. M. Richardson. Langmuir 9:242–248, 1993. G. Fragneto, R. K. Thomas, A. R. Rennie, and J. Penfold. Science 267:657–660, 1995. P. J. Atkinson, E. Dickinson, D. S. Horne, and R. M. Richardson. J. Chem. Soc. Faraday Trans. 91:2847–2854, 1995. B. Harzbollah, V. Angie´-Be´ghin, R. Douillard, and L. Bosio. Int. J. Biol. Macromol. 23:73–84, 1998. T. Nylander, F. Tiberg, T. J. Su, J. R. Lu, and R. K. Thomas. Biomacromolecules 2: 278–287, 2001.

21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43.

44. 45. 46. 47. 48.


Adsorbed Biopolymers in Food Applications 49. 50. 51. 52. 53. 54. 55. 56. 57. 58.

59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72.

73. 74. 75. 76. 77.

897

G. Fragneto, T. J. Su, J. R. Lu, R. K. Thomas, and A. R. Rennie. Phys. Chem. Chem. Phys. 2:5214–5221, 2000. M. A. Cohen Stuart, G. J. Fleer, J. Lyklema, W. Norde, and J. M. H. M. Scheutjens. Adv. Colloid Interface Sci. 34:477–535, 1991. G. J. Fleer, M. A. Cohen Stuart, J. M. H. M. Scheutjens, T. Cosgrove, and B. Vincent, Polymers at Interfaces. London: Chapman & Hall, 1993. F. A. M. Leermakers, P. J. Atkinson, E. Dickinson, and D. S. Horne. J. Colloid Interface Sci. 178:681–693, 1996. E. Dickinson, D. S. Horne, V. J. Pinfield, and F. A. M. Leermakers. J. Chem. Soc. Faraday Trans. 93:425–432, 1997. L. C. Ter Beek, M. Ketelaars, D. C. McCain, P. E. A. Smulders, P. Walstra, and M. A. Hemminga. Biophys. J. 70:2396–2402, 1996. D. S. Horne and J. Leaver. Food Hydrocolloids 9:91–95, 1995. J. Leaver and D. G. Dalgleish. Biochim. Biophys. Acta 1041:217–222, 1990. S. C. Russev, T. V. Arguirov, and T. D. Gurkov. Colloids Surf. B: Biointerfaces 19: 89–100, 2000. A. R. Mackie, J. Mingins, R. Dann, and A. N. North. In Food Polymers, Gels and Colloids (E. Dickinson, ed.). Cambridge: Royal Society of Chemistry, 1991, pp. 96– 112. Y. Fang and D. G. Dalgleish. J. Colloid Interface Sci. 196:292–298, 1997. M. Shimizu. In Food Macromolecules and Colloids (E. Dickinson and D. Lorient, eds.). Cambridge: Royal Society of Chemistry, 1995, pp. 34–42. J. A. DeFeijter, J. Benjamins, and M. Tamboer. J. Colloid Interface Sci. 90:289–292, 1982. R. J. Marsh, R. A. L. Jones, and J. Penfold. J. Colloid Interface Sci. 218:347–349, 1999. F. J. G. Boersboom, A. E. DeGroot-Mostert, A. Prins, and T. Van Vliet. Netherlands Milk Dairy J. 50:183–198, 1996. J. T. Petlov, T. D. Gurkov, B. E. Campbell, and R. P. Borwankar. Langmuir 16:3703– 3711, 1996. E. Dickinson and Y. Matsumura. Int. J. Biol. Macromol. 13:26–30, 1991. E. Dickinson, S. Rolfe, and D. G. Dalgleish. Food Hydrocolloids 3:193–203, 1989. S. Heńon and J. Meunier. Rev. Sci. Instrum. 62:936–939, 1991. D. Ho¨nig and D. Mo¨bius. J. Phys. Chem. 95:4590–4592, 1991. ˜ and M. Cejudo. J. Colloid J. M. Rodrı´guez Patino, S. C. Carrera, M. R. Rodrı´guez Nino, Interface Sci. 242:141–151, 2001. ˜ S. C. Carrera, and J. M. Rodrı´guez Patino. Colloids Surf. B M. R. Rodrı´guez Nino, 12:161–173, 1999. ˜ Food Hydrocolloids J. M. Rodrı´guez Patino, S. C. Carrera, and M. R. Rodrı´guez Nino. 13:401–408, 1999. ˜ and F. M. Cejudo. In J. M. Rodrı´guez Patino, S. C. Carrera, M. R. Rodrı´guez Nino, Food Colloids 2000: Fundamentals of Formulations (E. Dickinson and R. Miller, eds.). Cambridge: Royal Society of Chemistry, 2001, pp. 22–35. J. V. Boyd, J. R. Mitchell, L. Irons, P. R. Musselwhite, and P. Sherman. J. Colloid Interface Sci. 45:478–486, 1973. M. C. Phillips, M. T. A. Evans, D. E. Graham, and D. Oldani. Colloid Polym. Sci. 253:424–427, 1975. D. E. Graham and M. C. Phillips. J. Colloid Interface Sci. 70:427–439, 1979. J. R. Hunter, P. K. Kilpatrick, and R. J. Carbonell. J. Colloid Interface Sci. 142:429– 447, 1991. C. S. Gau, H. Yu, and G. Zografi. J. Colloid Interface Sci. 162:214–221, 1994.


898


78.

A. R. Mackie, A. P. Gunning, P. J. Wilde, and V. J. Morris. J. Colloid Interface Sci. 210:157–166, 1999. J. Benjamins, J. S. de Feijter, M. T. A. Evans, D. E. Graham, and M. C. Phillips. Faraday Discuss. 59:218–229, 1975. B. S. Murray and E. Dickinson. Food Sci. Technol. Int. 2:131–145, 1996. S. Damodaran. Adv. Food Nutr. Res. 34:1–79, 1990. P. Walstra. Chem. Eng. Sci. 48:333–349, 1993. J. Benjamins, Static and dynamic properties of proteins adsorbed at liquid interfaces. Ph.D. Dissertation, Wageningen University, Wageningen, The Netherlands, 2000. D. E. Graham and M. C. Phillips. J. Colloid Interface Sci. 70:403–414, 1979. F. MacRitchie. Adv. Protein Chem. 32:283–326, 1978. E. Tornberg. J. Colloid Interface Sci. 64:391–402, 1978. E. Tornberg. J. Sci. Food Agric. 29:762–776, 1978. A. V. Makievski, V. B. Fainerman, M. Bree, R. Wu¨stneck, J. Kra¨gel, and R. Miller. J. Phys. Chem. B 102:417–425, 1998. F. MacRitchie, Chemistry at Interfaces. San Diego: Academic Press, 1990, pp. 156– 185. A. V. Adamson, Physical Chemistry of Surfaces, 5th ed. New York: Wiley, 1990. S. S. Dukhin, G. Kretzschmar, and R. Miller, Dynamics of Adsorption at Liquid Interfaces. Theory, Experiment, Application. Amsterdam: Elsevier, 1995. ˜ Colloids Surf. A 103:91–108, 1995. J. M. Rodrı´guez Patino and M. R. Rodrı´guez Nino. ˜ In Food Macromolecules and J. M. Rodrı´guez Patino and M. R. Rodrı´guez Nino. Colloids (E. Dickinson and D. Lorient, eds.). London: Royal Society of Chemistry, 1995, pp. 103–108. ˜ Langmuir 15:4777– J. M. Rodrı´guez Patino, S. C. Carrera, and M. R. Rodrı´guez Nino. 4788, 1999. ˜ and S. C. Carrera. J. Agric. Food J. M. Rodrı´guez Patino, M. R. Rodrı´guez Nino, Chem. 47:2241–2248, 1999. G. Gonza´lez and F. MacRitchie. J. Colloid Interface Sci. 32:55–61, 1970. F. MacRitchie. Colloids Surf. 41:25–34, 1989. P. Joos, M. van Uffelend, and G. Serrien. J. Colloid Interface Sci. 152:521–533, 1992. ˜ and J. M. Rodrı´guez Patino. Ind. Eng. Chem. 2002 (in press). M. R. Rodrı´guez Nino D. E. Graham and M. C. Phillips. J. Colloid Interface Sci. 70:415–426, 1979. S. Damodaran and K. B. Song. Biochim. Biophys. Acta 954: 253–264, 1988. B. S. Murray, M. Faergemand, M. Trotereau, and A. Ventura. In Food Emulsions and Foams: Interfaces, Interactions, and Stability (E. Dickinson and J. M. Rodrı´guez Patino, eds.). Cambridge: Royal Society of Chemistry, 1999, pp. 223–235. M. Paulsson and P. Dejmek. J. Colloid Interface Sci. 150:394–403, 1992. A. F. H. Ward and L. Tordai. J. Chem. Phys. 14:453–461, 1946. ˜ C. Carrera, J. M. Navarro, G. RodriJ. M. Rodrı´guez Patino, M. R. Rodrı´guez Nino, guez, and M. Cejudo. Colloids Surf. B 21:87–99, 2001. L. G. Phillips, S. E. Hawks, and J. B. German. J. Agric. Food. Chem. 43:613–619, 1995. A. Prins. In Advances in Food Emulsions and Foams (E. Dickinson and G. Stainsby, eds.). New York: Elsevier Applied Science, 1988, pp. 91–122. J. N. de Wit and G. A. M. Swinkels. Biochim. Biophys. Acta 624:40–50, 1980. K. P. Das and J. E. Kinsella. J. Colloid Interface Sci. 139:551–560, 1990. E. Dickinson and S.-T. Hong, J. Agric. Food Chem. 42:1602–1606, 1994. M. Ahmed and E. Dickinson. Colloids Surf. 47:353–365, 1990. J. H. Crowe, L. M. Crowe, J. F. Carpenter, and C. A. Wistrom. Biochim. J. 242:1–10, 1987. J. C. Lee and S. N. Timasheff. J. Biol. Chem. 256:7193–7201, 1981.

79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93.

94. 95. 96. 97. 98. 99. 100. 101. 102.

103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113.


Adsorbed Biopolymers in Food Applications 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130.

131. 132. 133. 134. 135. 136.

137. 138. 139. 140. 141. 142.

143.

899

A. J. I. Ward and L. H. Regan. J. Colloid Interface Sci. 78:389–395, 1980. P. Chen, R. Prokop, S. S. Susnar, and A. W. Newman, Chapter 8. Proteins at Liquid Interfaces, Amsterdam: Elsevier, 1998. R. Miller and G. Kretzschmar. Adv. Colloid Interface Sci. 37:97–121, 1991. ˜ C. Carrera, M. Cejudo, and J. M. Rodrı´guez Patino. J. Am. Oil M. R. Rodrı´guez Nino, Chem. Soc. 78:873–879, 2001. G. L. Gaines, Insoluble Monolayers at Liquid–Gas Interface. New York: Wiley, 1996. ˜ and C. Carrera. AIChE J. 2001 (subJ. M. Rodrı´guez Patino, M. R. Rodrı´guez Nino, mitted). ˜ and C. Carrera. AIChE J. 2001 (subJ. M. Rodrı´guez Patino, M. R. Rodrı´guez Nino, mitted). B. S. Murray. In Proteins at Liquid Interfaces (D. Mo¨bius and R. Miller, eds.). Amsterdam: Elsevier, 1998, pp. 179–220. G. Kretzschmar and R. Miller. Adv. Colloid Interface Sci. 36:65–124, 1991. A. Prins. In New Physico-Chemical Techniques for the Characterization of Complex Food Systems (E. Dickinson, ed.). Glasgow: Blackie Academic, 1995, pp. 214–239. E. I. Franses, O. A. Basaran, and C. H. Chag. Curr. Opin. Colloid Interface Sci. 1: 296–312, 1996. M. Bos and T. van Vliet. Adv. Colloid Interface Sci. 91:437–471, 2001. J. Lucassen and M. van den Tempel. Chem. Eng. Sci. 27:1283–1291, 1972. L. Ting, D. T. Wasan, and K. Miyano. J. Colloid Interface Sci. 107:345–354, 1985. A. Bonfillon and D. Langevin. Langmuir 9:2172–2177, 1993. ˜ and M. Cejudo. LangJ. M. Rodrı´guez Patino, S. C. Carrera, M. R. Rodrı´guez Nino, muir 17:4003–4013, 2001. E. H. Lucassen-Reynders and J. Benjamins. In Food Emulsions and Foams: Interfaces, Interactions and Stability (E. Dickinson and J. M. Rodrı´guez Patino, eds.). Cambridge: Royal Society of Chemistry, 1999, pp. 195–206. ˜ P. J. Wilde, D. C. Clark, and J. M. Rodrı´guez Patino. Ind. Eng. M. R. Rodrı´guez Nino, Chem. Res. 35:4449–4456, 1996. K. Demetriades, J. N. Coupland, and D. J. McClements. J. Food Sci. 62:462–467, 1997. E. Dickinson and D. J. McClements, Advances in Food Colloids. London: Blackie Academic, 1995. F. J. Monahan, D. J. McClements, and J. B. German. J. Food Sci. 61:504–509, 1996. A. Ball and R. A. L. Jones. Langmuir 11:3542–3548, 1995. R. J. Green, I. Hopkinson, and R. A. L. Jones. In Food Emulsions and Foams: Interfaces, Interactions and Stability (E. Dickinson and J. M. Rodrı´guez Patino, eds.). Cambridge: The Royal Society of Chemistry, 1999, pp. 285–295. ˜ and C. Carrera. J. Agric. Food Chem. J. M. Rodrı´guez Patino, M. R. Rodrı´guez Nino, 47:3640–3648, 1999. G. R. Ziegler and E. A. Foegeding. Adv. Food Nutr. Res. 34:203–298, 1990. D. Oakenfull, J. Pearce, and R. W. Burley. In Food Proteins and Their Applications (S. Damodaran and A. Paraf, eds.). New York: Marcel Dekker, 1997, pp. 111–142. J. I. Boye, C.-Y. Ma, and V. R. Harwalkar. In Food Proteins and Their Applications (S. Damodaran and A. Paraf, eds.). New York: Marcel Dekker, 1997, pp. 25–56. T. Beveridge, L. Jones, and M. A. Tung. J. Agric. Food Chem. 32:307–313, 1984. J. E. Kinsella, D. J. Rector, and L. G. Phillips. In Protein Structure–Function Relationships in Food (R. Y. Yada, R. L. Jackman, and J. L. Smith, eds.). London: Blackie Academic, 1994. M. Verheul, S. P. F. Roefs, J. Mellema, and K. G. deKruif. Langmuir 14:2263–2268, 1998.


900


144.

M. Verheul, S. P. F. Roefs, and K. G. deKruif. J. Agric. Food Chem. 46:896–903, 1998. J. A. Hunt and D. G. Dalgleish. J. Food Sci. 42:2131–2135, 1994. J. A. Hunt and D. G. Dalgleish. J. Food Sci. 50:1120–1123, 1995. T. Nylander and M. N. Wahlgren. Langmuir 13:6219–6225, 1997. E. Dickinson, R. H. Whyman, and D. G. Dalgleish. In Food Emulsions and Foams (E. Dickinson, ed.). London: Royal Society of Chemistry, 1987, pp. 40–51. H. Casanova and E. Dickinson. J. Agric. Food Chem. 46:72–76, 1998. E. Dickinson, M. G. Semenova, and A. S. Antipova. Food Hydrocolloids 12:227–235, 1998. V. N. Ismailova, G. P. Yamployskaya, and B. D. Tulovskaya. Colloids Surf. A Physicochem. Eng. Aspects 160:89–106, 1999. S. O. Agboola and D. G. Dalgleish. J. Food Sci. 60:399–404, 1995. S. O. Agboola and D. G. Dalgleish. Lebensm. Wiss. Technol. 29:425–432, 1996. A. Kulmyrzaev, R. Chenamai, and D. J. McClements. Food Res. Int. 33:15–20, 2000. E. Dickinson and C. M. Woskett. In Food Colloids (R. D. Bee, P. Richmond, and J. Mingins, eds.). Cambridge: Royal Society of Chemistry, 1989, pp. 74–96. J. Leaver, D. S. Horne, and A. J. R. Law. Int. Dairy J. 9:319–322, 1999. L. Leaver, A. J. R. Law, and D. S. Horne. In Food Emulsions and Foams: Interfaces, Interactions and Stability (E. Dickinson and J. M. Rodriguez Patino, eds.). Cambridge: Royal Society of Chemistry, 1999, pp. 258–268. J. Leaver, A. J. R. Law, and E. Y. Brechany. J. Colloid Interface Sci. 210:207–214, 1999. D. S. Horne, P. J. Atkinson, E. Dickinson, V. J. Pinfield, and R. M. Richardson. Int. Dairy J. 8:73–77, 1998. A. R. Mackie, A. P. Gunning, P. J. Wilde, and V. J. Morris. Langmuir 16:2243–2247, 2000. C. M. Wijmans and E. Dickinson. Langmuir 15:8344–8348, 1999. ˜ Langmuir 15:2484– J. M. Rodrı´guez Patino, S. C. Carrera, and M. R. Rodrı´guez Nino. 2492, 1999. ˜ and J. M. Rodrı´guez Patino. Colloids Surf. B S. C. Carrera, M. R. Rodrı´guez Nino, 12:175–192, 1999. ˜ S. C. Carrera, and M. Cejudo. J. Colloid J. M. Rodrı´guez Patino, M. R. Rodrı´guez Nino, Interface Sci. 240:113–126, 2001. ˜ S. C. Carrera, and M. Cejudo. LangJ. M. Rodrı´guez Patino, M. R. Rodrı´guez Nino, muir 2001 (in press). A. R. Mackie, A. P. Gunning, M. J. Ridout, P. J. Wilde, and J. M. Rodrı´guez Patino. Biomacromolecules 2:1001–1006, 2001.

145. 146. 147. 148. 149. 150. 151. 152. 153. 154. 155. 156. 157.

158. 159. 160. 161. 162. 163. 164. 165. 166.


Biopolymers at Interfaces, Second Edition (Surfactant Science)

Biopolymers at Interfaces, Second Edition (Surfactant Science Series)

Biopolymers at Interfaces, Second Edition